Fundamentals of Optical Fibre

2.
Optical Fibers
2.1 Fundamentals of Optical Fibers
The term optical fibers indicates special forms of optical waveguides, the most im-
portant special features of which are:
¾rotationally symmetrical cross-section
¾flexible
¾can be produced in great lengths
The characteristics of optical fibers are determined by a multitude of possible
constructive details. For example, the material selected primarily determines the
attenuation and the thermal stability. On the other hand, the optical bandwidth, in
essence the transmission capacity, is determined by the refractive index profile.
This is most likely the reason why most optical fibers are named after their index
profile. All current variations will be presented in the following sections.
2.1.1 Refractive Index Profiles
The properties of wave guiding through a fiber are governed largely by the profile
of the refractive index of the core and cladding. In a step index profile fiber the
refractive index is constant across the entire cross section of the core and cladding
(Fig. 2.1) while the light rays propagate along straight lines in the core and are
completely reflected at the core/cladding interface.
r
a
refractive
index n(r)
-a
Fig. 2.1: Refractive index profile in a step index profile fiber

38 2.1 Fundamentals of Optical Fibers
The profile of the refractive index in the core and in the cladding is expressed as
follows:
n(r ) ncore for r d a
n(r ) ncladding for r ! a
where a is the core radius.

The individual rays cover different distances, so that there are considerable dif-
ferences in their respective transit times. Choosing a fiber with a graded-index
profile can minimize these differences. Fibers with a graded-index profile are
made up of a core having a radius-dependent refractive index and a cladding with
a constant refractive index (Fig. 2.2):
§ g
§r· ·
n(r )² ncore, max ² ¨1 ' ¨ ¸ ¸ for r d a
¨ © a ¹ ¸¹
©
n(r ) ncladding for r ! a
where g is the profile exponent

and ' is the relative refractive index difference:
ncore 2 ncladding 2
'
2 ncore 2
r
a
-a
refractive
index n(r)
Fig. 2.2: Principle of a fiber with a graded-index profile
Those rays propagating in the center travel a shorter distance, but because of
the higher refractive index there, they travel at a lower speed. On the other hand,
the smaller refractive index near the cladding causes the rays traveling there to
have a higher velocity, but they have a longer distance to travel. By choosing a
suitable profile exponent it is possible to compensate for these differences in tran-
sit time. For negligible chromatic dispersion the ideal profile exponent is 2. One
then speaks of a parabolic index profile.
2.1 Fundamentals of Optical Fibers 39
2.1.2 Numerical Aperture
When light enters the fiber's input face at an angle 4max, it is refracted at an angle
Dmax (Fig. 2.3). Applying the law of refraction we have:
n0 sin4max ncore sinD max ncore sin(90 - J max )
n0 sin4max ncore cos J max
n0 sin4max
ncore 1 - sin2 J max , with ncladding ncore 2 sin2 J max
n0 sin4max
ncore 1 - ncladding ncore 2
n0 sin 4max ncore 2 ncladding 2 , for n0 1 follows
sin 4max ncore 2 ncladdimg 2
The sine of the maximum incident-ray angle 4max is defined as the numerical
aperture AN (Fig. 2.3). The angle 4max is referred to as the acceptance angle, and
twice the acceptance angle is referred to as the aperture angle. Using the relative
refractive index difference ', the value for AN is obtained as:
AN sin 4max ncore 2 '
n0
Jmax
4max
Dmax
Fig. 2.3: Definition of the acceptance angle
Thus, the value of the numerical aperture (NA) is solely dependent on the diffe-
rence in the refractive indices of the core/cladding material.
Example: The refractive indices of a standard PMMA fiber are ncore = 1.49 and
ncladding = 1.40; we thus obtain AN = 0.50 and 4max = 30q.
Whereas the numerical aperture of the step-index profile fiber remains constant
over the entire core, the graded-index profile fiber exhibits a decreasing accep-
tance angle from the center of the core to the cladding (Fig. 2.4).
Fig. 2.4: Acceptance angle of a graded-index profile fiber
Compared with other fiber types (Fig. 2.5), POF has the largest numerical
aperture and the largest core diameter. This is one of the most important advan-
tages of POF, since the connection technology that can be used for POF is more
economical to apply than that used for glass fibers.
singlemode glass fiber multimode glass fiber polymer fiber

10/ 125 μm 50/ 125 μm 980/ 1000 μm
multimode glass fiber 62.5/ 125 μm

(plastic clad)
200/ 230 μm
100/ 140 μm
0 mm 0.5 mm 1.0 mm
Fig. 2.5: Aperture angle and core diameter of glass fibers and polymer fibers
2.1.3 Ray Trajectory in Optical Fibers
In the step index profile fiber, light propagates along a zigzag path, being totally
reflected at the core/cladding interface; in the graded-index profile fiber, light pro-
pagates on a sinusoidal trajectory that is created within the graded-index profile
through refraction. If the incident light rays lie within one and the same plane
through which the fiber axis runs, meridional rays are formed. In all other cases,
skew rays are formed. Figure 2.6 shows the projection onto the fiber's incident
face. Step and graded-index profile fibers show the same behavior. The speci-
fication of the numerical aperture always refers to the meridional rays.
D’
D’
D’
D’
Fig. 2.6: Meridional rays
Skew rays form an angle of \ < 90q with the tangential plane at the
core/cladding interface (Fig. 2.7). They never cross the fiber axis and propagate
along screw-like paths. For step index profile fibers, the projection onto the cross-
sectional area resembles a polygonal line so that these rays do not cross a circle-
shaped area having a radius rk around of the axis.
rk \
Fig. 2.7: Skew rays in step index profile fibers
In graded-index fibers with a parabolic profile, ellipses are formed in the pro-
jection (Fig. 2.8 left) that may under certain circumstances form circles; these rays
are called helical rays (Fig. 2.8 right). Their distance from the fiber axis is always
constant.
Fig. 2.8: Helical rays (left) and skew rays (right) in graded-index profile fibers
2.1.4 Modes in Optical Fibers
2.1.4.1 The Mode Concept

The phenomena of refraction and reflection discussed so far can be graphically
explained with the help of geometrical optics, whereby the size of the wavelength
and the diameter of the finite ray are not considered (O and dray= 0). However, to
obtain a complete description of the wave guiding phenomenon, the wave pro-
perties of light must also be considered. The goal is to calculate the electric field
and intensity distribution of the light in the optical fiber. The Eigenvalue equation
is derived and solved on the basis of the Maxwell equations. Ref. [Blu98] provides
a detailed description. The solutions to the Eigenvalue equations are a finite
number of field distributions within the light waveguide. These field distributions
are referred to as modes of the waveguide. If we apply this concept to the ray
model, this means that apparently not all incidental rays for which 4 < 4max is true
can propagate, but rather only those rays that have a particular angle. Figure 2.9
illustrates this situation: in order for light to propagate in a particular direction, a
wave must constructively overlap itself with its own reflecting wave in such a way
that the phase position is repeated after double reflection. The black lines
perpendicular to the direction of propagation identify the planes with the same
phase angle. The spacing is O/ncore.
O/ncore transverse direction
electrical field
Fig. 2.9: Formation of the mode structure within the waveguide
Whereas the zigzag paths would lead to intensity distributions within the ray-
optical model that would change depending on the length of the fibers, the wave
model provides a constant light-dark distribution that is independent of the length
across the waveguide's cross-section.
The number N of the guided modes is approximately described by:
1 g
N| V2
2 g2
where V = 2S a AN/O, a is the radius and g is the profile exponent (see also Sec-
tion 1.1.5).
For step-index profiles g o f. This results in a value of N | ½ · V² for the
number of modes. For parabolic profiles g = 2 and thus N | ¼ · V². A polymer op-
tical fiber with AN = 0.5, a core radius of 0.5 mm and a wavelength of O = 650 nm
can carry 2.9 million modes. If the angle of total reflection is exceeded, radiation
modes are created and the light is radiated into the cladding. If the refractive index
of the cladding is higher than the surrounding medium (air, for example), cladding
modes may be formed. In the POF, the optical cladding is encased in an absorbing
jacket so that no cladding modes can form. In contrast to guided modes, it is not
possible to count radiation modes. They do not take part in signal transmission.
(Fig. 2.10 special conditions for POF are explained below). Higher modes propa-
gate under a larger angle, lower modes under a smaller one. Under certain circum-
stances skew rays may turn into so-called leaky waves, which, on the one hand,
are guided in the Z-direction and, on the other hand, transfer energy to the
cladding. Under certain conditions they can still be detected in POF even after
several 10s of meters. Hence, they can influence both the transmission process as
well as the measuring techniques used.
radiation mode
higher mode
lower mode
cladding mode
Fig. 2.10: Guided, cladding and radiation modes
The following equation describes the relationship between the angles D, \ and
G in Fig. 2.11 ([Sny83]):
cos D sin G sin ȥ
D is the angle of the incident and reflected ray relative to the surface normal of
the tangential plane in P. \ describes the angle between the reflection plane and
the tangent plane, and G is the angle between the projection of the skew ray on the
cross-sectional plane and the direction of propagation (parallel to the fiber axis).
Figure 2.12 summarizes the various ray types according to the respective angles
derived from the above equation ([Bun99a]). For guided rays holds G < Gmax and D
> Dmax. The leaky waves are shown in the subsequent rectangle while the ray
modes are shown above the line D = Dmax. For meridional rays D = 90q - G because
\ = 90q, i.e. they lie on the blue line.
D
D G
\
P
Fig. 2.11: Designation of the angles of a skew ray; the right diagram shows the angle G,
which is obtained by projecting the skew ray on to the cross-sectional plane
0 \ q
D[°] inside this triangle there
10 are the radiation rays
20
30 inside this triangle there \ q
are the guided rays
40 \ q
50
meridional \ q
60 rays
70
Dmax
inside this rectangle there
80
Gmax are the leaky modes
G[°]
90
0 10 20 30 40 50 60 70 80 90
Fig. 2.12: The different types of rays
2.1.4.2 Mode Propagation in Real Fibers
Several chapters of this book discuss the special characteristics of light propa-
gation in POF. Here now, the processes that need to be considered will be looked
at as a whole. The function of fibers as a waveguide for passing on light by means
of total reflection at the core/cladding interface has already been discussed.
If the ray model were applied consistently, then a light ray launched into an
ideal fiber would always propagate at the same angle relative to the fiber axis.
With a divergent light source, the far field would always remain constant along the
length of the fiber. This would not be true for the near field, as Fig. 2.13 illus-
trates: depending on the course of the ray, different locations along the fiber would
generate different near fields in the form of point structures. However, this contra-
dicts the results obtained through experiments: there a continuous distribution of
intensity is obtained, and from a certain length onwards the intensity does not
change at all. Although the ray model is very illustrative, its practical application
is limited as the example above shows. In order to be able to describe experi-
mental results it is thus necessary to move on to the mode concept. In this respect
it is important to keep in mind that many optical simulation programs work on the
basis of discrete light rays. In order to obtain truly realistic results, a sufficient
number of rays has to be simulated.
only a few launched discrete modes
nearfield (very schematically)
Fig. 2.13: Near fields under conditions of the ray model with only a few discrete light
paths (in practice very difficult to measure and visible only on very short
lengths)
2.1.5 Parameters for Describing Real Fibers and Waveguides
In order to describe the characteristics of real fibers and waveguides different

parameters are defined which vary in importance depending on their respective
application. All of these parameters are influenced by the propagation conditions
of the different modes. In the case of multimode fibers most characteristics depend
typically on the mode distribution. This means that a fiber initially allows the
propagation of light in different paths (modes). Depending on the light sources at
the front end of the fiber, not all these modes are launched, at least not with a uni-
form power distribution. Since each mode has different characteristics, an altered
behavior of the fiber is on average the result. In addition, the problem becomes
more complicated since an exchange of energy between the modes can occur over
the length.
Typical fiber characteristics will be defined and explained in the following
sections. Uniform Mode Distribution (UMD) and Equilibrium Mode Distribution
(EMD) are the usual standard measuring conditions.
2.1.5.1 Attenuation
The most important process encountered by light as it passes through a fiber is at-
tenuation. When passing through an optical fiber of the length L, the power of the
light decreases (Fig. 2.14). The following equation applies to the optical power:
PL P0 e DcL
where PL and P0 are the power of the light after passage through a fiber of
length L in km and at the front end of the fiber, respectively; D´ is the value of the
attenuation coefficient in km-1.
P0 PL
Fig. 2.14: Definition of attenuation
To make it easier to work with the numbers involved here, it is usual to express
attenuation logarithmically. Thus, the attenuation coefficient is expressed as D in
dB/km.
10 P
D log 0 4,343 Dc
L PL
Attenuation value a is the non-dimensional variable (given as a number or in

dB) obtained from the product D · L. Figure 1.19 illustrates the relationship bet-
ween the attenuation value and the change in power as a percentage.
attenuation factor a [dB]

30 25 20 15 10 5 0
0.1 1 10 100
power ratio PL/P0 [%]
Fig. 2.15: Conversion of the power ratio PL/P0 in % into the dB value
Very often there is not a clear differentiation in the technical literature between
attenuation per unit length D and attenuation factor a. One often speaks simply of
the attenuation of the fiber. The addition “spectral” refers to the wavelength
dependence. A mistake is avoided, however, when the unit is indicated. We still
have to mention that attenuation and attenuation per unit length are practically
always indicated as positive numbers.
Quantity Symbol Unit Formula

attenuation coefficient, lin. D´ km-1 {ln (P0/PL)}/L
attenuation coefficient, log. D dB/km {10log (P0/PL)}/L
attenuation a dB 10log (P0/PL)
Especially in the area of optical short-range communication, indicating the fiber

attenuations in dB is much more practical than, for example, representing the ab-
solute transmission. POFs are being used more and more in the near infrared range
for quite short transmission lengths. Finally, PMMA can also be used for wave-
guide structures in the mm range. Fig. 2.16 shows the attenuation curve of a
PMMA-POF according to [Hess04].
100,000
attenuation [dB/km]
30,000
10,000
3,000
1,000
300 theory
measured
100
30
wavelength [nm]
10
500 600 700 800 900 1000
Fig. 2.16: Attenuation spectrum of the PMMA-POF (theory and measured by [Hess04])
Nevertheless, the representation comprises approximately 3 decades, i.e. a fac-

tor of 1,000 which cannot be overlooked on a linear scale.
2.1.5.2 Mode-Dependent Attenuation

When talking about glass fibers, it is often assumed that the attenuation of all light
rays is identical. For practical purposes, this assumption is sufficiently accurate.
With POF, the path difference between the rays parallel to the axis and the propa-
gation directions close to the critical angle of total reflection can become quite
large. For the standard NA-POF with AN = 0.50 this difference is about 6%. For
polycarbonate fibers with AN = 0.90, the difference is even 21%. For this reason
alone, there is a considerably greater level of attenuation where large propagation
angles are involved. In 100 m of POF, a light ray of this type will travel 6 m
farther which results in an additional loss of more than 1 dB when the attenuation
level is 200 dB/km. At 1,000 dB/km for polycarbonate fiber, this would result in
an additional loss of 4 dB after 20 m of travel (less than 50% of the launched
power reach the fiber output).
The second, more significant cause for mode-dependent attenuation is the
attenuation resulting from the cladding material. Fluorinated polymers are used as
optical cladding for PMMA fibers; these claddings may have an attenuation of
several 10,000 dB/km [Paar92]. Locking more exactly on the propagation of a
plane wave at the interface, we find that, even if total reflection results, the electri-
cal field escapes into the optically thinner medium by a distance in the order of
magnitude of the wavelength. This process is also known as the Goos-Hänchen
Shift ([Bun99a]) and the model explains this as resulting from a shift of the reflec-
tion plane into the optically thinner medium. The reflected ray is hence slightly
displaced on the interface surface, as can be seen in Fig. 2.17. In this model, the
additional light path would be subjected to the higher attenuation of the cladding
material.
cladding
core
area of higher
attenuation
Fig. 2.17: Goos-Hänchen shift
Although the light path in the cladding is only in the Pm range for each reflec-
tion, it still plays a significant role because of the much higher attenuation encoun-
tered there. This effect is particularly striking when the core diameters are reduced
in size. Theoretically speaking, attenuation and bandwidth should not be depen-
dent on the core diameter. Nevertheless, thin cores such as those used in multi-
core fibers have indeed considerably larger bandwidths [Tesh98], a slightly
increased attenuation and narrower far-field widths. These effects are explained
quite well in [Bun99b] and [Ziem99c].
This effect also occurs in glass fibers. Silica glass fibers with a polymer
cladding (PCS) have losses in the core below 10 dB/km (wavelength range from
650 nm to 1,300 nm), whereas the polymer cladding has an attenuation of several
100 to 1,000 dB/km.
Attenuation values of 180 dB/km for the core and 9,000 dB/km for the cladding
are indicated in [Ebb03] for step index profile glass-glass fibers (used in fiber
bundles). Reasonably priced conventional glasses - albeit much purer than in win-
dow glass - are used in these fibers and not silica glass.
In singlemode and graded-index profile silica fibers there are no mentionable
differences in attenuation between the core and the cladding since both consist of
Si02. The germanium dopant in the core does not have any great influence. An
important consequence of the mode-dependent attenuation is, as will be discussed
later on, a significantly narrower far field after greater fiber lengths than one
would expect from the fiber NA.
2.1.5.3 Mode Coupling

The term mode coupling refers to the process by which energy from one direction
of propagation is transferred to several others. This can happen, for example at
scattering centers. Since the light scattering in a PMMA-POF makes up a con-
siderable part of the attenuation, this process is always present. Figure 2.18
clarifies the procedure (still in the ray model).
scattering center
Fig. 2.18: Mode coupling at a scattering center
Many experimental results clearly indicate that mode coupling occurs predomi-
nately at the core/cladding interface (Fig. 2.19). This can be explained by the fact
that is it not possible to create an ideal surface in the sub-nanometer range when
very large polymer molecules are involved. Thus, mode coupling is also depen-
dent on the angle of propagation.
cladding
core scattering
center
Fig. 2.19: Mode coupling at the core/cladding interface
Mode coupling alters the bandwidth of a fiber. When collimated light is

launched, energy is gradually transferred to the higher angle ranges so that mode
dispersion increases and bandwidth decreases. If light is introduced in all angle
ranges, so that maximum differential delays occur, energy is exchanged between
the angles so that the initially slower rays become “faster” and vice versa. Accor-
ding to the laws of statistics, the differential delay (or more precisely, the standard
deviation) does not increase in a linear relationship to the length but approxi-
mately only proportional to the square root of the length. This applies to lengths in
excess of a characteristic coupling length, which for PMMA-POF is generally
several 10 m.
Mode coupling always results in additional attenuation. Whenever there are
changes in the light propagation, energy is coupled into those angle ranges in
which there is no longer any light guiding. The shorter the coupling length, the
larger the additional attenuation will be. If the observed behavior of the POF,
namely the filling up of the near field after a few 10 cm of fiber, could be ex-
plained exclusively because of the mode coupling, then additional attenuations in
the range of 1000 dB/km would result - which indeed does not occur.
Figure 2.20 shows an electron microscope picture of the core-cladding interface
layer (photo ZWL, 2003). The marked smooth part running from the top left to the
bottom right is the surface of the core with the cladding removed. At the top right
you can see the cracked core. The step is the 10 μm thick optical cladding. Further
theoretical considerations on the problems of scattering can be found in [Kru06a]
and [Kru06b].
Fig. 2.20: Photo of the core-cladding interface of SI-POF taken by electron microscope
(ZWL Lauf)
2.1.5.4 Mode Conversion

The definition of propagation angles or of modes actually applies only to wave-
guides that are straight. It takes just one bend to make a different approach neces-
sary. The most precise method would be to recalculate the modes for the system of
the now bent fiber; however, this is theoretically and practically much too com-
plex a process. It is more appropriate to consider the zone before and after the
bend as a straight waveguide and, at the bend, to perform a transformation onto
the new reference axis. Formally, light is thus transmitted from one propagation
direction to another, as Fig. 2.21 demonstrates.
fiber axis behind

fiber axis in front of a bend
a bend
new propagation angle
Fig. 2.21: Mode conversion at a bend
Strictly speaking, mode conversion can be described as a special case of mode

coupling. The difference is that the number of modes or the propagation directions
is not increased. In the POF mode conversion most likely occurs at the core/clad-
ding interface surface, for example at micro bends or at fluctuations in the re-
fractive index difference. The question of the influence of mode conversion and
coupling on the additional attenuation depends essentially on the angle depen-
dency of the processes. The more the direction of the light is altered, the more los-
ses occur. A quantitative analysis of these processes for POF is extremely difficult
and is yet to be carried out. However, for the physical processes assumed, mode
coupling should have a larger angle-independent contribution (scattering on larger
inhomogenities).
Fig. 2.22: Far fields of different POF (product A/B at the top/bottom); left/right after
20 m/50 m of fiber, launch with collimated light (AN Launch < 0.016)
An impressive experiment that confirmed this statement was shown in [Poi00].

If collimated light is launched into a SI-POF, a ring-shaped far field can be gene-
rated at the output even after 50 m of fiber, for which purpose the fiber might be
properly bent. This experiment can only be explained under the assumption that
mode conversion predominates. However, the different fibers made by different
manufacturers show considerable differences in their behavior which do not
necessarily have an effect on attenuation.
It is easy to see here that the mode field is not completely filled even after 20 m
to 50 m (Fig. 2.22).
2.1.5.5 Mode Coupling Lengths

The length of a fiber in which a state of equilibrium arises through mode conver-
sion and coupling is described as coupling length whereby different definitions
exist. The best known is the description with the aid of a length-dependent band-
width. Here the coupling length is the point at which the linear decrease in the
bandwidth turns to a root dependency (see Fig. 2.36). In practice this point is diffi-
cult to measure, but other parameters such as far field width and attenuation,
change with fiber length. For example, values for the kilometric attenuation with
different launch conditions are shown In Figures 2.23 and 2.24.
400 D [dB/km]
fiber “A”
350 source “I”
source “II”
300 source “III”
source “IV”
250
200
150
100
lPOF [m]
50
1 2 5 10 20 50 100
Fig. 2.23: Attenuation of a SI fiber under different excitation (acc. to [Lub02b])
Both diagrams show very clearly that the different launch conditions (source I
emits very widely, source IV nearly collimated) lead to extremely different attenu-
ation values. After some ten meters, however, the differences disappear for the
most part through mode coupling. Evidently, there are great differences among the
fiber types.
D [dB/km]
400
fiber “B”
350
source “I”
source “II”
300
source “III”
source “IV”
250
200
150
100
lPOF [m]
50
1 2 5 10 20 50 100
Fig. 2.24: Attenuation of another SI-fiber at different launch conditions
The next two figures 2.25 and 2.26 show measurements of far field widths for a
POF and a PCS each with altered launch conditions. Once again it can clearly be
seen how the differences caused by the different coupling conditions are evened
out after some 10 to 100 m.
far field width [°] NALaunch: 0.64 0.33 0.09

32
0.48 0.19 0.05
30
28
26
24
22
20
18
16
14 PMMA SI-POF
12 POF length [m]
10
5 20 50 100
Fig. 2.25: Launch dependent far field widths of a PMMA SI-POF
In the 200 μm thick PCS it takes considerably longer to establish the equili-
brium mode distribution especially when the length is related to the fiber diameter.
The values of the NA (calculated from the 5% far field width) are represented for
lengths up to 500 m.
0.40
measured NA
0.35
0.30 AN = 0.02
AN = 0.09
0.25 AN = 0.17
AN = 0.26
0.20 AN = 0.34
AN = 0.48
0.15
fiber length [m]
0.10
1 10 100 1000
Fig. 2.26: Excitation dependent far field width of a 200 μm-PCS
In general, the mode coupling length is characterized as the distance in which a

parameter has come closer by 1/e to the state of equilibrium. For example, this
corresponds to the charging time constants of a capacitor. One cannot therefore
say that EMD conditions exist after one coupling length. Depending on how large
the tolerated deviations are, several coupling lengths have to be considered.
Figure 2.27 shows the theoretical curve of a parameter.
mode dependent fiber parameter [a.u.]

600
500
characteristic at Lc = 100 m
400
parameter
deviation for parameter
300
short fibers deviation for
1, 2 und 3 u Lc
200
equilibrium mode value

100
10 20 50 100 200 500 1000
fiber lenght [m]
Fig. 2.27: Approximation of an optical parameter to the equilibrium value by mode coup-
ling (schematically)
2.1.5.6 Leaky Modes
The significance of leaky modes has already been touched upon earlier. For the
sake of completeness, it should be noted here again that light rays that lie above
the critical angle of the total reflection do not entirely vanish but still contribute
significantly to light propagation even after several 10s of meters.
Not until we examine the interaction of attenuation, mode-dependent attenu-
ation, mode coupling and mode conversion and take leak modes into account, can
we establish a model for the light propagation of SI polymer fibers that can at least
qualitatively describe the experimentally observed behavior. In principle, the same
processes take place in GI-POF; however there are basic differences:
¾ With GI-POF, there is no core/cladding transition to serve as an essential
cause for mode coupling, mode conversion, and mode-dependent attenuation.
¾ Fluorinated GI-POF are used in wavelength ranges in which Rayleigh
scattering is less significant.
¾ To form the index profile, various zones of the fiber, as seen from the axis,
are provided with varying concentrations of a dopant or a copolymer so that
the attenuation usually gets a gradient. This is probably the most significant
cause of mode-dependent attenuation in GI-POF.
Yabre and Zubia made comprehensive observations on mode propagation in
GI-POF [Yab00a], [Yab00b], [Arr99], [Arr00].
The problem of mode coupling and mode conversion is sure to be very inte-
resting for multi step index fibers. Bandwidths could result that are larger than
what is theoretically expected. Some different theoretical investigations were
made in cooperation between the POF-AC and the University of Bilbao (Spain).
More details will be given in the fiber simulation chapter.
As the example of the multi-core fibers shows, mode-dependent attenuation can
be used to exchange attenuation for bandwidth. Less attenuating cladding would
reduce the overall attenuation of the POF, but more than likely also reduce the
bandwidth (always assuming equilibrium mode distribution). The future will
decide which parameter is of greater significance for users. If the transmission
budget is sufficiently large, it would be possible to increase the bit rate though
multi-level coding or by electrically compensating the dispersion so that a reduc-
tion in attenuation is the minimum goal to be targeted in this field.
2.1.5.7 Dispersion in Optical Fibers

Dispersion refers initially to all processes that result in a difference in the transit
times of various modes. One mode is thereby always a propagation condition of
the light that is uniquely defined by the wavelength, polarization, and propagation
path.
Differential delays between the various light components lead to a reduction in
the modulation amplitude of higher frequencies. This makes the fiber a low-pass
filter.
The bandwidth of a fiber communication transmission system is usually con-

sidered to be the frequency for which the optical level of a sine-modulated signal
has dropped by 3 dB. Strictly speaking this approach only applies to a Gaussian
low-pass filter. This means that a pulse of insignificant width will correspond to
the Gauss function after it has traveled the length of the fiber:
P( f ) P0 ( f ) e

- f 2 f02
where P(f) is the power of a random frequency f at the end of the measuring
path, P0(f) is the launched power and f0 is a constant that describes the bandwidth.
Figure 2.28 illustrates the process schematically.
a)
P0(f)
b) pulse response
c)
d)
e)
P(f)
time t
Fig. 2.28: Effect of dispersion on a sine-wave signal
Curve 'a' shows the sine-modulated source optical signal (it must be noted that
optical power can only take positive values). Figure 'b' shows how a single pulse
approaching a Gaussian function after traveling through the fiber. This is a theore-
tical borderline case because the Gaussian function extends from -f to +f, but the
output pulse cannot begin before the input pulse has started. To measure the shape
of the complete output signal, the input signal can be split into a series of pulses,
as shown in Fig. 'c'. After traveling through the fiber, every pulse forms a Gaus-
sian function of the respective height (Fig. 'd'). These have to be brought together
again to achieve the result in curve 'e' (mathematically speaking, this is a convolu-
tion of the input pulse with the so-called pulse response of the transmission link).
It is easy to see that the amplitude of the signal has decreased. Attenuation of
the light has not been taken into consideration.
A short light pulse is briefly broadened when it travels the length of a fiber
(Fig. 2.29) and this in turn reduces the transmission bandwidth.
optical optical
input output
power power
100 %
50 % 100 %
optical
50 %
fiber
time time
tin tout
Fig. 2.29: Pulse broadening by passing an optical fiber
If Gaussian-shaped pulses are assumed, the result of the pulse broadening 't is
the square root of the difference of the squares of the input and output pulse width
(FWHM full width at half maximum):
2 2
't t out t in
The consequence of this broadening is that the time gap between the bits
becomes smaller, that the pulses finally overlap and that the receiver can no longer
differentiate between the two. The transmission bandwidth is limited as the light
waveguide functions as a low-pass filter. The product of bandwidth and length
characterizes the transmission capacity of a fiber. [Gla97] applies to Gaussian-
shaped pulses:
0.44
B L | L
't
Pulse broadening is caused by mode dispersion and chromatic dispersion. For
multimode fibers it is necessary to consider the factors of material, modes and pro-
file dispersion (in graded index fibers). Waveguide dispersion additionally occurs
in singlemode fibers, whereas profile dispersion and mode dispersion do not.
All the kinds of dispersion appearing in optical fibers are summarized in
Fig. 2.30. The mechanisms dependent on the propagation paths are marked in
yellow, whereas the wavelength-dependent processes are marked in green.
dispersion
modal dispersion chromatic dispersion

(multimode fibers) (multimode and singlemode fibers)
profile material dispersion

dispersion (multimode and
(multimode fibers) singlemode fibers)
polarization mode dispersion waveguide dispersion

(singlemode fibers) (singlemode fibers)
Fig. 2.30: Dispersion mechanisms in optical fibers
In regard to the fibers and applications dealt with in this book only mode and
chromatic (material) dispersion play a role so that the following sections deal
solely with these two effects.
2.1.5.8 Mode Dispersion

Since the light paths have different lengths, the pulses that have started simultane-
ously arrive at different times at the fiber's output, a fact that leads to pulse broa-
dening. Figure 1.29 shows the 'fastest' (D = 0) and the 'slowest' (D = Dmax) rays.
ncladding
Jmax
L2 a
2
Dmax
1
L1
ncore
Fig. 2.31: Deriving the difference in the transit time

The propagation times of the two different propagation paths are determined
purely geometrically for:
n
t1 L1 core
c
2
n L1 ncore 1 L1 ncore
t2 L 2 core
c c sin J max c ncladding
n § ncore ncladding ·
' t mod t 2 t1 L1 core ¨ ¸
c ¨ n cladding
¸
© ¹
L1 2 L1 ncore
AN | '
2 c ncladding c
Figure 2.32 shows the dependence of the bandwidth on the numerical aperture
with which the light is launched. The assumption is that the far field, i.e. the
angular distribution of the light in the fiber, will remain constant over the entire
length of the sample (no modal coupling or conversion). For a PMMA standard
fiber with an AN = 0.50, a differential delay of 't | 25 ns for 100 m is produced.
The transit time is proportional to the square of the NA. From the above-men-
tioned expression B | 0.44/'tmod, a value of 15 MHz results for the bandwidth.
theoretical bandwidth [MHz]

1,000
500
200 fiber-
length:
100 10 m
50
25 m
20 50 m
75 m
10 100 m
0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60
numerical aperture
Fig. 2.32: Bandwidth calculated as a function of the launch NA

The critical angle Jmax of total reflection is determined by the ratio of both
refractive indices (example, 1.492 for the core and 1.456 for the cladding):
1.456
J = arcsin = arcsin 0.976 = 77.4q
1.492
(max. angle to axis : Dmax 12.6q)
Thus, the relationship between both paths y and z is:

z = y/sin (D) = y 1.0247
The NA of this fiber is determined by:

0.5 0.5
AN = (ncore2 - ncladding2) = (1.4922 -1.4562) = 0.32
The pulse broadening for a fiber length L is derived as follows:

Transit time of the parallel-axis modes: t1 = L · n/c0
Transit time of the modes with max. angle: t2 = L · n/c0 · 1.0247
Differential delay: 't = L · n/c0 · 0.0247
For example for 100 m, n = 1.492: 't = 12.3 ns
With the approximation B · 't = 0.44 B = 33 MHz
Different NA lead to different bandwidths, whereby a doubling of the NA

reduces the bandwidth to a quarter:
Theoretical bandwidth: AN = 0.60: 10 MHz 100 m
AN = 0.50: 14 MHz 100 m
AN = 0.40: 22 MHz 100 m
AN = 0.30: 40 MHz 100 m
AN = 0.25: 57 MHz 100 m
AN = 0.19: 97 MHz 100 m
To correctly calculate the theoretical bandwidth, it is just not sufficient to con-
sider the two possible ray paths selected here. A very comprehensive description
of mode propagation in POF is provided in [Bun99a]. In the ray model, each
possible propagation direction is described by the two angles D and G (for an
explanation of these angles, please refer to Fig. 2.11).
As far as transit time is concerned, only angle G is of relevance. Figure 2.33 is
an illustration from [Bun99a] of the zone of the guided rays and the leaky rays
which themselves are again subdivided. Regardless of the size of \, angle G
cannot exceed a particular maximum value so that a maximum possible differen-
tial delay is the consequence.
Only the marked triangle contains not attenuated rays that are capable of propa-
gation. If one assumes that all possible propagation paths have the same energy
(UMD - uniform mode distribution), it can be seen that paths having a larger
propagation angle are more probable than rays traveling parallel to the axis.
D>q@ \ q
70
\ q
meridional \ q
75
rays
80 \ q
Gmax
85
guided
rays
90 G>q@
0 5 10 15 20
Fig. 2.33: Possible rays in an optical fiber
As measurements of the far field (that is the power as a function of the angle to
the fiber axis, measured in a sufficient large distance) of a POF shows, this is also
reflected in the greater power obtained with larger angles. If the power is ex-
pressed in relation to the solid angle element, a constant power density is found
because larger angles cover a correspondingly larger arc. This is shown schemati-
cally in Fig. 2.34.
rel. power/solid angle

1.0
UMD
0.8
0.6
0.4
0.2
0.0 angle to the

-30 -20 -10 0 10 20 30 fiber axis [°]
Fig. 2.34: Power distribution with UMD
The differential delay increases approximately by the square of the angle rela-
tive to the fiber axis. If a short pulse having a mode distribution that correspon-
ding to UMD is launched into the fiber input, an approximately rectangular pulse
is generated at the output of the length of which corresponds to the approximate
values shown above for the maximum differential delay. Figure 2.35 demonstrates
the precise results for an assumed attenuation-free standard NA POF for the pulse
form obtained after 10 m, 20 m, 50 m, and 100 m of ideal POF (from [Bun99a]).
norm. signal
100%
80%
60%
10 m 20 m 50 m 100 m
40%
20%
time [ns]
0%
0 5 10 15 20 25 30 35
Fig. 2.35: Output pulses of a POF under UMD conditions ([Bun99a])
Real SI-POF provide considerably higher bandwidths. The main reason for this
is the presence of mode-dependent attenuation in conjunction with mode mixing,
as will be shown in the next chapter.
The differential delay 't increases proportionally to a particular length Lc
(coupling length); for longer lengths, the increase is sub-linear (Fig. 2.36). The
following holds true:
't v L for L L c
't v LN for L ! L c with N 1
whereby the exponent N must be determined for each fiber. It is typically bet-
ween 0.5 and 0.7. The coupling length Lc ranges between 30 m and 40 m for stan-
dard SI-POF.
2.5
pulse broadening [a.U.] W a l
2.0
1.5 Wal
in reality
1.0
0.5
Lc
length [m]
0.0
0 20 40 60 80 100 120 140 160 180 200
Fig. 2.36: Schematically representation of the pulse broadening reflecting mode coupling
effects
The impulse response of a 50 m long standard POF can be seen in Fig. 2.37.
The half-value width of the impulse amounts to about 50 ns, i.e. only about 30%
of the expected value. Furthermore, it is noticeable that the rear pulse edge drops
more slowly. It is in this range that the higher modes lie which are attenuated very
greatly by the mode-dependent losses. The dropping off of the rising edge can be
explained by the effect of modal mixing.
0.7
U [V] theoretical
0.6 pulse shape
'W = 16 ns
0.5
0.4
0.3
'W = 5 ns
0.2
0.1
0.0
0 5 10 15 20 25 30 35 40
t [ns]
Fig. 2.37: Real pulse shape for 50 m St.-NA-POF
Calculating the bandwidth of graded-index fibers is clearly more complex.

Current studies in this field can be found in [Yab00a], [Yab00b] and [Arr99].
Profile dispersion occurs in graded-index profile fibers. It is the remainder of
the mode dispersion that can no longer be compensated for and it depends on the
relative refractive index difference ', which in turn is wavelength-dependent. An
optimization of the profile exponent can be accomplished for a certain wavelength
for which d'/dO = 0. A profile exponent of g | 2 causes a temporal broadening of:
L1 ncore '2
't prof ,
c 2
in other words, a factor '/2-reduced broadening of the pulse as compared with
step index POF; for a typical graded-index POF this means a reduction by
approximately 2 orders of magnitude [Blu98]. Mode dispersion or profile disper-
sion can only be avoided by using singlemode fibers. As explained later on, due to
the combination with the chromatic dispersion, certain polymer fibers, have some
advantages as opposed to silica glass fibers.
2.1.5.9 Chromatic Dispersion

Chromatic dispersion describes the influence of the spectral width of a transmitter
on a temporal broadening of the input pulse. This includes the material-dispersion
and waveguide-dispersion types of dispersion. Both effects also occur in single-
mode fibers. Waveguide dispersion is caused by the fact that light waves penetrate
into the fiber cladding to various depths, depending on the wavelength of the light
wave. Thus, the different speeds of the core and cladding parts result in pulse
broadening. Since only a small portion of the light wave in higher modes of large
diameter fibers spreads into the cladding, this effect is only considered for single-
mode fibers.
However, even if only one mode is allowed to propagate, pulse broadening
occurs due to material dispersion. Every light source has a spectral width 'O > 0.
The following applies for the pulse broadening 'tmax due to material dispersion:
O d² nO
' t mat L 'O L 'O MO
c d O²
where 'O is the spectral width of the transmitter

n(O): wavelength-dependent refractive index,
M(O): material dispersion parameters usually given in ps/kmnm
Figure 2.38 shows the influence of material dispersion on pulse broadening,
using polymer fibers as an example. Corresponding to the material dispersion, the
longer wavelengths (red) propagate with a greater velocity than the shorter ones
(blue).
output pulse
't
input pulse time
length
'O
wavelength
Fig. 2.38: Temporal broadening as a result of material dispersion
The real influence of the chromatic dispersion from different polymer optical
fibers to the system bandwidth will be shown in the next chapter which will
contain detailed descriptions of the materials and fiber types.
2.2 Index Profiles and Types of Fibers 65
2.2 Index Profiles and Types of Fibers
After the theoretical descriptions on the properties of optical fibers in the section
on the fundamentals of light propagation and the observations indicated above on
mode propagation and the essential characteristics of fibers this following section
will deal with concrete, available fibers. First, the different index profiles, as
briefly mentioned in 1.1.6, will be introduced using examples.
The next section shows the historical development especially in regard to the
different POF variants. Thereafter the important characteristics attenuation and
bandwidth will be shown in a series of experimental results.
Three parameters are basically responsible for the actual properties of optical
fibers. The core and cladding materials used determine the attenuation and chro-
matic dispersion. The refractive index profile determines the mode dispersion and
the core diameter is also responsible for the number of modes. Especially the core
material and the index profile are at least recognizable from the name of the fiber,
a designation method widely used in this book.
In the following section the historical development of the different polymer
fibers is summarized. The POFs are dealt with in regard to their index profiles.
Thereafter, different hybrid and glass fibers for short-range data transmission will
also be introduced. The following chapter deals especially with the bandwidth of
thick optical fibers since this characteristic is particularly important and also it
makes the greatest demands on measurement techniques.
2.2.1 Step Index Profile Fibers (SI)
As was the case with silica glass fibers, the first polymer optical fibers were pure
step index profile fibers (SI-POF). This means that a simple optical cladding sur-
rounds a homogenous core. For this reason a protective material is always in-
cluded in the cable. Figure 2.39 schematically represents the refractive index
curve.
As already shown above, the refractive index step determines the numerical
aperture (NA) and thus the acceptance angle. Some typical values are shown in
Table 2.1. The refractive index of the core was always taken as 1.5, whereas the
cladding has a correspondingly smaller refractive index. The last line is valid for
wave guiding against air (n = 1). Here an acceptance angle of 90° is valid since the
NA exceeds the value of 1.
ncore
ncladd
jacket core jacket

optical cladding optical cladding
Fig. 2.39: Structure of a step index profile fiber
66 2.2 Index Profiles and Types of Fibers
Table 2.1: Relationship between relative refractive index difference and numerical aper-
ture (core refractive index = 1.50)
Relative Refractive- Refractive Index Numerical Acceptance Angle

Index-Difference of the Cladding Aperture of the Fiber
0.22 % 1.497 0.10 6°
0.4 % 1.494 0.13 8°
0.8 % 1.488 0.19 11°
1.0 % 1.485 0.21 12°
1.5 % 1.478 0.26 15°
2.0 % 1.470 0.30 17°
2.7 % 1.460 0.35 20°
4.0 % 1.440 0.42 25°
5.8 % 1.413 0.50 30°
8.0 % 1.380 0.59 36°
12.0 % 1.320 0.71 45°
20.0 % 1.200 0.90 64°
33.3 % 1.000 1.12 90°
A larger acceptance angle of the fiber simplifies the launching of light, e.g.
from a semi-conductor source. In addition, a high NA reduces the losses asso-
ciated with fiber bending, as schematically illustrated in Fig. 2.40.
rays, exceeding the

critical angle of total
reflection behind the
bend
launched light
rays
bend
rays, guided
radius
behind the
bend
Fig. 2.40: Loss at fiber bends
Due to the effects of bending, the propagation direction of each individual ray
is changed relative to the axis of the fiber. In the case of multi-mode fibers, a part
of the rays is always extracted because the rays exceed the angle of total reflection
at the interface between core and cladding. For fibers with a large NA, the effect
of a change in angle for a certain amount of bending is not so significant so that
the bending losses diminish. Likewise, when coupling fibers to each other (at
connectors) the loss due to angle errors is less significant when there is a large
numerical aperture.
A disadvantage of fibers with a large NA is the greater difference in time delay
between the different light paths, and this in turn leads to a greater level of mode
dispersion. This limits the bandwidth. In addition, the loss at coupling points in-
creases if there is a gap between the abutting faces. Some advantages of larger or
smaller numerical apertures are listed in Table 2.2.
Table 2.2: Influence of higher NA to various fiber parameters

Property of the Fiber Behavior with increasing NA
bending sensitivity becomes smaller
fiber coupled power becomes higher
connecting loss for fiber angular mismatch becomes smaller
connecting loss for axial fiber gap becomes higher
connecting loss for fiber axis lateral gap becomes higher
bandwidth becomes smaller
Silica glass multi-mode fibers usually have an NA of approximately 0.20. Silica

glass fibers with polymer cladding have an NA in the range of 0.30 to 0.40 (some-
times 0.65). The large refractive index difference between the materials that are
used for the core and the cladding of polymer fibers allows significantly higher
NA values. The majority of the initially produced SI-POF had an NA of 0.50 (e.g.
[Asa96], [Esk97], [LC95]). SI-POF with an NA around this value are nowadays
generally called standard NA-POF or standard POF for short. The bandwidth of
such fibers is approximately 40 MHz for a 100 m long link (quoted as the band-
width-length product 40 MHz · 100 m). For many years this was a completely
satisfactory solution for most applications.
2.2.2 The Step Index Fiber with Reduced NA (low-NA)
However, when it became necessary to replace copper cables with polymer optical
fiber to accomplish the transmission of ATM data rates of 155 Mbit/s (ATM:
asynchronous transfer mode) over a distance of 50 m, a higher bandwidth was
required for the POF. In the mid-nineties all three important manufacturers deve-
loped the so-called low-NA POF.
POF with a reduced numerical aperture (low-NA POF) feature a bandwidth
increased to approximately 100 MHz · 100 m because the NA has been reduced to
approximately 0.30. The first low-NA POF was presented in 1995 by Mitsubishi
Rayon ([Koi98]). Figure 2.41 shows that the fiber construction corresponds to the
standard POF, the distinction being that the refractive index difference is smaller
(approximately 2 %). Usually the same core material is used, but the cladding
material has a modified composition.
ncore
ncladding
jacket core jacket

Fig. 2.41: Structure of a low-NA step index profile fiber
Unfortunately, practical testing showed that although this fiber met the require-
ments of the ATM forum ([ATM96b]) with respect to bandwidth, it did not meet
the requirements with respect to bending sensitivity. These requirements specify
that for a 50 m long POF link the losses resulting from a maximum of ten 90°
bends having a minimum bending radius of 25 mm should not exceed 0.5 dB. In
order to meet both these requirements at the same time it became necessary to find
a new structure.
2.2.3 The Double-Step Index Optical Fiber (DSI)
The double-step index POF features two claddings around the core, each with a
decreasing refractive index (Fig. 2.42). In the case of straight installed links, light
guiding is achieved essentially through the total reflection at the interface surface
between the core and the inner cladding. This index difference results in an NA of
around 0.30, similar to the value of the original low-NA POF.
ncore
ncladding1
ncladding2
jacket core jacket
outer / inner inner / outer

Fig. 2.42: Structure of a double step index profile fiber

When fibers are bent, part of the light will no longer be guided by this inner
interface. However, it is possible to reflect back part of the decoupled light in the
direction of the core at the second interface between the inner and the outer
cladding. At further bends, this light can again be redirected so that it enters the
acceptance range of the inner cladding. The inner cladding has a significantly
higher attenuation than the core. Light propagating over long distances within the
inner cladding will be attenuated so strongly that it will no longer contribute to
pulse propagation. Over shorter links the light can propagate through the inner
cladding without resulting in too large a dispersion. A schematic illustration is
shown in Fig. 2.43.
launched light rays
rays behind
the bend
1 rays, guided only by 3

the inner cladding
1
2 rays, guided by the
outer cladding behind
the bend 2
bend 1
3 rays, guided by the radius
outer cladding over a 2
limited distance
4 not guided rays
behind the bend
Fig. 2.43: Operation of a bent double step index profile fiber
The first generation of DSI-POF primarily served the purpose of increasing the
bandwidth of 1 mm fibers from 40 MHz · 100 m to 100 MHz · 100 m with an un-
changed minimum bending radius of 25 mm. The respective applications are to be
found in LANs and home networks.
The fiber producers offer these fibers under the same type names as the original
“real” low-NA fibers. It has since become standard procedure to call the fibers
low-NA and to indicate DSI as the index profile.
Currently, another goal is being pursued: the bandwidth of standard POF is
sufficient for applications in vehicle networks, but the bending radius should be
reduced. Presently being discussed are POFs, the index steps of which correspond
to a NA of 0.50 or 0.65 respectively to the inner and outer cladding. The bending
radius can thus almost be halved.
2.2.4 The Multi-Core Step Index Optical Fiber (MC)
As described above, the requirements of high bandwidth and low sensitivity to

bending are difficult to accomplish together within one and the same fiber having
a diameter of 1 mm. Fibers with a smaller core diameter can solve this problem
since the ratio to the fiber radius is larger for the same absolute bending radius.
However, this contradicts the requirements for easy handling and light launching.
A PCS with a core diameter of 200 ȝm and an AN = 0.37 permits, for example, a
bending radius of 5 mm with very low bending losses.
As a compromise, Asahi developed a multi-core fiber (MC-POF, see [Mun94],
[Mun96] and [Koi96c]). In this fiber many cores (19 to over 200) are put together
in production in such a way that together they fill a round cross-section of 1 mm
diameter.
First, the individual fibers are all perfectly round and each has its own optical
cladding. Only a certain share of the total cross-section of the bundle enters the
cores guiding the light, since the cladding areas and the spaces between the fibers
have to be accounted for. Figure 2.44 shows the parameters which mark the per-
centage of the filled-in area. The number N here indicates how many fibers lie
next to each other over a diameter while n indicates the entire number of fibers.
R R
dm
dm
r
N=5
N=1 n = 19
Fig. 2.44: Schematically arrangement of cores in a MC-POF
In the figure, R denotes the radius of the complete fiber (typically 0.5 mm) and
d the thickness of the optical cladding (e.g. 5 μm). Let us assume first of all that
the individual cores are arranged in a hexagonal shape with N = 2z + 1 cores
positioned next to each other.
The next Fig. 2.45 shows how the arrangement for fibers is changed for
z = 1 to 5. While these sketches can give a clear definition of the number of fibers
that can be arranged within a circular shape, for smaller and smaller individual
cores the possibilities are more complex. The arrangement at the bottom right
shows one possible deviation. For the first five arrangements the number of indi-
vidual fibers is calculated as follows:
n = 3z2 + 3z + 1.
It follows that the individual radius r is:

r = R/N = R/(2z + 1).
N=3 N=5 N=7
N=9 N=11 N´=11
Fig. 2.45: Possible circular arrangements of cores in a MC-POF
In Table 2.3, the degree of coverage of the circle area is calculated for the cases
shown. First, the number of individual cores is calculated from z. The radius r
results from the overall radius of the fiber (here always 500 μm). Parameter ta indi-
cates what percentage of the total circular area is covered by the individual circles
(for the hexagonal arrangement of an infinite number of circles a maximum of
90.69 % of the area can be covered). When calculating parameter tb, the fact that
part of the cross-section is lost to the optical claddings (all uniformly 5 μm thick)
is taken into account.
Table 2.3: Core cross area degree of coverage for MC fibers (ideal)
z: N: n: r: ta: tb:
0 1 1 500 μm 100.00 % 98.01 %
1 3 7 167 μm 77.78 % 73.18 %
2 5 19 100 μm 76.00 % 68.59 %
3 7 37 71.4 μm 75.51 % 65.31 %
4 9 61 55.6 μm 75.31 % 62.36 %
5 11 91 45.5 μm 75.21 % 59.57 %
11´ 85 49.3 μm 82.47 % 66.57 %
6 13 127 38.5 μm 75.15 % 56.88 %
7 15 169 33.3 μm 75.11 % 54.27 %
8 17 217 29.4 μm 75.09 % 51.73 %
14 29 631 17.2 μm 75.03 % 37.82 %
f - - - 90.69 % -
Figure 2.46 shows the proportion of core area tb as depending on the number of
cores for four different thickness’ of the optical cladding.
use of the total cross area tb

100%
dm = 5 μm dm = 20 μm
dm = 10 μm dm = 30 μm
80%
60%
40%
20%
0% number of
1 7 19 37 61 91 127 169 217 single cores
Fig. 2.46: Proportion of core area for different cladding thickness
As can be expected, the proportion of the overall covered area decreases with
an increasing number of cores because the proportion of cladding area will be-
come larger and larger. A certain minimum thickness of cladding is necessary for
it to be able to fulfill its function and still be technologically feasible. The four
individual data points show the case of the optimized fiber arrangement with 85
individual cores in accordance with Fig. 2.45.
Given a minimum thickness of the optical cladding between 5 μm and 10 μm,
these considerations indicate that a maximum number of some 100 single cores
should be used, in which case the proportion of useable area will hardly exceed
70 %. It is easy to conclude that a smaller proportion of useable core area would
lead to an increase in the losses encountered when connecting transmitters to, and
fibers between each other.
Fig. 2.47: 37 core POF with deformed single cores (schematically)

Practical experience shows that a better utilization of the area can be achieved.
During the manufacturing process the fibers are placed together at higher tempe-
ratures which means that they change their shape and thus reduce the gaps bet-
ween the fibers. Apparently, the resulting deviations from the ideal round shape do
not play a significant role in light propagation (the causes for this are not yet
completely understood; some points worth discussing can be found in the chapter
on light propagation in POF). Figure 2.47 shows a schematic illustration of the
cross-section of a fiber with 37 cores, such as e.g. in [Tesh98]. Data of available
MC-POF and -GOF are grouped together later.
Figure 2.48 shows the refractive index profile of a MC-POF, shown as a cross-
section through the diameter of the fiber. The index steps correspond to those of a
standard POF.
ncore
ncladding
jacket cores jacket
Fig. 2.48: Structure of a step index multi core fiber
Since the bandwidth only depends on the NA for SI fibers, it should be possible
to measure values comparable to the standard POF. However, the fact is that the
measured values are actually significantly higher, which has been explained in the
chapter 2.1.5.2 discussing mode-selective attenuation mechanisms.
Glass fibers are also produced for use in many areas as fiber bundles. In
lighting technology fiber glass bundles with a large NA are widely spread. (The
lighting of the headlight outer ring at BMW via such a fiber bundle is well-
known.) In the meantime, such fibers are also available for data communication
([Lub04b]).
2.2.5 The Double Step Index Multi-Core Fiber (DSI-MC)
In the MC-POF, too, an increase in bandwidth was achieved by reducing the index
difference. Due to the smaller core diameters it was still possible to avoid an
increase in bending sensitivity.
Even better values were achieved with individual cores having a two-step
optical cladding such as illustrated in Fig. 2.49. The principle is the same as in the
double-step index POF with an individual core. In this case a bundle with single
cladding is completely surrounded by a second cladding material (“sea/islands”
structure).
ncore
ncladding1
ncladding2
jacket cores jacket

outer / inner optical cladding
Fig. 2.49: Structure of a double step index profile multi core POF
2.2.6 The Graded Index Optical Fiber (GI)
When using graded index profiles (GI) an even greater bandwidth becomes pos-
sible. In these profiles, the refractive index continually decreases (as a gradient),
starting from the fiber axis and moving outwards to the cladding. Of particular
interest are profiles that follow a power law (remember chapter 1.4.1).
ª g
§ distance to fiber axis · º
refractive index n=n «1- ' ¨ ¸ »
fiber axis «¬ © core radius ¹ »¼
The parameter g - often also Į - is characterized as the profile exponent. When

g = 2 one speaks of a parabolic profile. The borderline case of the step index
profile fibers corresponds to g = . The parameter ǻ states the relative refractive
index difference between the maximum core and the cladding refractive index.
Figure 2.50 shows a parabolic index profile.
ncore
ncladding
jacket core jacket

Fig. 2.50: Structure of a graded index profile fiber
Due to the continually changing refractive index, the light rays in a GI fiber do
not propagate in a straight line but are constantly refracted towards the fiber axis.
Light rays that are launched at the center of the fiber and do not exceed a certain
angle are completely prevented from leaving the core area without any reflections
occurring at the interface surface. This behavior is illustrated schematically in
Fig. 2.51. The geometric path of the rays running on a parallel to the axis is still
significantly smaller than the path of rays that are launched at a greater angle.
However, as can be seen, the index is smaller in the regions distant from the
core. This means a greater propagation speed. In an ideal combination of para-
meters the different path lengths and different propagation speeds may cancel each
other out completely so that mode dispersion disappears. In reality, this is only
possible in approximation. It is possible, however, to increase bandwidths by two
to three orders of magnitude compared with the SI fiber.
step index profile fiber graded index profile fiber
n n
Fig. 2.51: Comparison of step and graded index profile (see also chapter 2.1.1)
When considering not only the pure mode dispersion but also chromatic disper-
sion, i.e. the dependence of the refractive index on the wavelength and spectral
width of the source, an optimum index coefficient 'g' deviating from 2 is achieved.
This has been the subject of comprehensive investigations by the research group
around Prof. Koike ([Koi96a], [Koi96b], [Ish00], [Koi97a], [Koi96c], [Koi98] and
[Ish98]). In [Ish00] and [Koi00] the significance of this effect is particularly pro-
nounced (see also Chapter 2). Due to the smaller chromatic dispersion of fluori-
nated polymer compared with silica, the bandwidth of GI-POF theoretically achie-
vable is significantly higher than that of multi-mode GI silica glass fibers. In parti-
cular, this bandwidth can be realized over a significantly greater range of wave-
lengths. This makes the PF-GI-POF interesting for wavelength multiplex systems.
However, in this case the index profile must be maintained very accurately, a
requirement for which no technical solution has as yet been provided.
Another factor involved in the bandwidth of GI-POF is the high level of mode-
dependent attenuation ([Yab00a]) compared to silica glass fibers. In this case
modes with a large propagation angle are suppressed resulting in a greater band-
width. An example is the simulation that was carried out in [Yab00a]: the band-
width of a 200 m long PMMA-GI-POF increases from 1 GHz to over 4 GHz,
taking into account the attenuation of higher modes. This is also confirmed in
practical trials. Mode coupling is less significant for GI fibers than it is for SI
fibers since the reflections at the core-cladding interface do not occur.
2.2.7 The Multi-Step Index Optical Fiber (MSI)
Following the many technological problems experienced in the production of gra-

ded index fibers having an optimum index profile that remains stable for the dura-
tion of its service life, an attempt was made to approach the desired characteristics
with the multi-step index profile fiber (MSI-POF). In this case the core consists of
many layers (e.g. four to seven) that approach the required parabolic curve in a
series of steps. Here a “merging” of these steps during the manufacturing process
may even be desirable. A diagram of the structure is shown in Fig. 2.52.
ncore
ncladding
jacket core jacket
Fig. 2.52: Structure of a multi step index profile fiber
In this case light rays do not propagate along continually curved paths as in the
GI-POF, but on multiple diffracted paths as demonstrated in Fig. 2.53. However,
given a sufficient number of steps, the difference to the ideal GI profile is relati-
vely small so that large bandwidths can nevertheless be achieved. MSI-POF were
presented in 1999 by a Russian institute (Tver near Moscow [Lev99]) and by
Mitsubishi (ESKA-MIU, see [Shi99]). In the meantime, other companies are pro-
ducing such fibers which are often called GI fibers. These GI and MSI fibers are
classified in the same class of standards, e.g. A4e.
Fig. 2.53: Light propagation in the MSI-POF
2.2.8 The Semi-Graded Index Profile Fibers (Semi-GI)
A relatively new version of index profiles are fibers which have a gradient with a
slightly varying index above the core cross section, but do have an optical clad-
ding with a great index step as shown in Fig. 2.54 ([Sum00], [Sum03], [Ziem05f]
and [Ziem06i]).
ncore
ncladding
jacket core jacket
Fig. 2.54: Structure of a semi-graded index profile fiber
At first sight this variety of fiber has enormous advantages. Light which propa-
gates within the gradient is only subject to very little mode dispersion. If a ray of
light has a greater propagation angle, e.g. after being bent, then it continues to be
led to the core-cladding interface layer through total reflection. However, these
rays do have a very much higher mode dispersion. Figure 2.55 shows how light
spreads theoretically and what consequences this has for the pulse response.
input output
„GI“-modes
„SI“-modes
t
Fig. 2.55: Light propagation in semi-graded-index profile fibers theoretically
In principle, two different groups of modes can be seen in the picture. The
paths designated as GI modes do not touch the cladding and only show a very
slight difference in propagation times. The shares designated as SI modes are
completely reflected at the core-cladding interface layer. These light paths are also
bent in the core, but the light path, now very much longer, can no longer be com-
pensated for in the outer areas by the lower refractive index. With very high data
rates the second mode group is drawn out so widely that it is presented solely as a
kind of DC offset in the eye diagram. At the POF-AC a data rate of 1 Gbit/s was
transmitted over 500 m of a GI PCS fiber with a PRBS signal ([Vin05a]). Data
rates up to 3 Gbit/s could be attained with a small surface APD receiver ([Kos95]).
In order to do justice to the complex behavior of the semi-GI POF, corresponding
modulation formats should be selected.
2.2.9 An Overview of Index Profiles
Figures 2.56 through 2.58 again show all index profiles described in an overview.
Due to the wide range of possibilities offered in polymer chemistry further deve-
lopments are certainly to be expected. For example, multi-core graded fibers,
fibers with special cladding for a reduction of the losses at the core/cladding inter-
face or to increase the bandwidth or even multi-core fibers with different indivi-
dual cores are all conceivable. In the following figures POF variants are shown
with typical parameters.
SI-POF Low-NA-POF DSI-POF

AN = 0.50 AN = 0.30 AN = 0.30
40 MHz100 m 100 MHz100 m 100 MHz100 m
Fig. 2.56: POF with single core and step index profile
Single-core fibers with diameters between 125 μm and 3 mm are available from
different manufacturers at a reasonable price and in robust quality. Most of the
polymer optical fibers used in practical applications are of these types.
MC-SI-POF MC-DSI-POF
e.g. 200 cores e.g. 37 cores
AN = 0.30 AN = 0.19
100 MHz100 m 400 MHz100 m
Fig. 2.57: POF with multiple cores and step index profile
MC fibers are available from various manufacturers. They are deployed in

applications ranging from high data rates transmission systems through to optical
image guides. Because of the short lengths produced, the prices are still signifi-
cantly above expectations. However, further developments in this field can be ex-
pected in the future.
GI-POF MSI-POF
AN = 0.20 AN = 0.30
2 GHz100 m 500 MHz100 m
Fig. 2.58: Polymer fibers with graded index and multi step index profile
Graded index as well as multi-step index profile POF are commercially avai-
lable today. Laboratory experiments and a series of practical installations in Japan
and Europe, (e.g. [Mös04]) show the great potential in regard to the bit rates
possible. Asahi Glass introduced them into the market around 2001. Lucent Tech-
nologies, later called OFS and trading under the name of Chromis Fiberoptics as
of 2004 ([Whi04], [Park05a]), also announced the possibility of producing large
amounts of GI POF in case of demand.
2.3 The Development of POF 79
In Europe, fibers by Nexans are manufactured in Lyon ([Gou04]). All three fi-
bers will consist of the fluorinated polymer material CYTOP®. The core diameter
of the LucinaTM Fiber by Asahi Glass is 120 μm with an AN = 0.28. A protective
cladding made from PMMA and measuring 500 μm is placed around an area of
fluorinated polymer outside the core profile. The duplex cable has external dimen-
sions of approximately 3 by 5 mm. The lowest attenuation achieved to date is ap-
prox. 15 dB/km for a wavelength of 1,300 nm. The specified value is < 50 dB/km
for 700 nm - 1,300 nm.
There has also been significant progress in the manufacture of GI or MSI-POF
respectively on a PMMA basis (see Section 2.3.4).
2.3 The Development of Polymer Optical Fibers
The following sections will describe the polymer fibers presented so far, whereby
particular attention will be paid to the chronological sequence of the develop-
ments. Section 2.4 supplements these observations with some types of multimode
glass fibers which were not discussed in the first edition.
2.3.1 Looking back
The first POF were manufactured by DuPont as early as the late sixties. Due to the
incomplete purification of the monomer materials used, attenuation was still in the
vicinity of 1,000 dB/km. During the seventies it became possible to reduce losses
nearly to the theoretical limit of approximately 125 dB/km at a wavelength of
650 nm. At that point in time glass fibers with losses significantly below 1 dB/km
at 1,300 nm/1,550 nm were already available in large quantities and at low prices.
Digital transmission systems with a high bit rate were then almost exclusively
used in telecommunications for long-range transmissions. The field of local com-
puter networks was dominated by copper cables (either twisted-pair or coaxial)
that were completely satisfactory for the typical data rates of up to 10 Mbit/s com-
monly used then. There was hardly any demand for an optical medium for high
data rates and small distances so that the development of the polymer optical fiber
was slowed down for many years. A significant indicator for this is the fact that at
the beginning of the nineties the company Höchst stopped manufacturing polymer
fibers altogether.
During the nineties, after data communication for long-haul transmission had
become completely digitalized, the development of digital systems for private
users was commenced on a massive scale. In many spheres of life we are being in-
creasingly confronted with digital end user equipment. The CD player has largely
replaced analog sound carriers (vinyl records and cassettes). The MP3 format is
leading to a revolution in music recording and distribution. The DVD (Digital
Video Disc) and large hard disk drives could lead to the replacement of the analog
video recorder within a few years. Even today more digital television programs
80 2.3 The Development of POF
are available than analog programs. Decoder boxes have become standardized
(MPEG2 format) and will be integrated into television sets in the future. More and
more households are using powerful PC and digital telephone connections
(ISDN). With offers such as T-DSL (ADSL technology provided by Deutsche
Telekom AG) as well as fast internet access via satellite or broadband digital ser-
vices on the broadband cable network, private users are being offered access to
additional digital applications even before the start of the new millennium.
Likewise, in the automotive field the step towards digitalization has long been
made. CD changers, navigation systems, distance-keeping radar and complex con-
trol functions are increasingly part of the standard equipment being provided in all
classes of vehicles. The development of electronic outside mirrors, fast network
connections even from within an automobile as well as automatic traffic guidance
systems will ensure a further increase in the range of digital applications for the
motor vehicle. All these examples demonstrate that completely new markets for
digital transmission systems are being developed for short-range applications.
Polymer optical fibers can meet many of these requirements to an optimum degree
and are therefore increasingly of interest.
A significant indicator for this development is the history of the International
Conference for Polymer Optical Fibers and Applications which has been taking
place annually since 1992 and represents the most significant scientific event in
this specialized field. Many of the developments described below were presented
for the first time at these conferences.
2.3.2 Step Index Polymer Fibers
The SI-POF is the oldest variant of all polymer fibers. Its development goes back
to the beginning of the 1960’s, i.e. in a period when silica glass fibers were being
developed. Today the SI-POF is by far the most common POF variant. In Table
2.4 data from different publications on this fiber type are summarized - without
claiming to be complete.
Table 2.4: Published data of SI-POF

Ref. Year Producer Product Øcore Attenuation at O NA Remarks
μm dB/km nm
[Min94] 1963 Du Pont CROFON - 1.000 650 st. first POF
[Koi97a] 1964 Du Pont - 500 650 st.
[Koi96c] 1968 Du Pont - 500 650 st. first SI-POF
[Sai92] 1976 Mitsubishi Eska - 300 650 st.
[Min94] 1978 Mitsubishi Super Eska - 300 650 st.
[Koi95] 1982 NTT - 55 568 st.
[Sai92] 1983 Mitsubishi Eska Extra - 124 650 st. 4 MHzkm
[Sai92] 1983 Mitsubishi Eska Extra - 65 570 st.
[Koi95] 1983 Mitsubishi 1000 110 570 st.
[Min94] 1984 Mitsubishi Eska Extra - 150 650 st.
Table 2.4: Published data of SI-POF (continued)

Ref. Year Producer Product Øcore Attenuation at O NA Remarks
μm dB/km nm
[Koi95] 1985 Asahi - 80 570 st.
[Sai92] 1991 Mitsubishi Eska Extra - 125 650 st. up tp 85°C
[Sai92] 1991 Mitsubishi Eska Extra - 65 570 st.
[Koi95] 1991 Hoechst 1000 130 650 st.
[Tesh92] 1992 Asahi Luminous-F - 175 660 0.50 310 MHz10m
AN, LED=0.50, 105°C
[Tesh92] 1992 Asahi X-1 - - - 0.37 540 MHz10m
AN, LED = 0.50
[Tesh92] 1992 Asahi X-2 - - - 0.28 >1.000 MHz10m
AN, LED = 0.50
[Eng96] 1992 Höchst EP51 970 190 650 st. 90 MHz100 m
with 650 nm LED
[Kit92] 1992 Mitsubishi Eska Premier 1000 135 650 0.51 up to 85°C
[Lev93] 1993 CIS Sveton MN- 200- 150 650 0.45 up to 70°C
Series, Grade U 600
[Lev93] 1993 CIS Sveton MF- 200- 120 650 0.48 up to 70°C
Series, Grade U 1000
[Non94] 1994 Sumitomo n. a. 480 150 650 0.51 200 MHz50m
'n=0.055
[Koe98] 1998 Mitsubishi n. a. 1000 110 650 0.47 80 MHz100 m
[Mye02] 2002 Dig. Optr. n. a. 1000 - - 0.50 2003 announced
[Luv03] 2003 Luvantix SI type 1000 160 650 0.40 200 MHz bandwidth
[Nuv04] 2004 Nuvitech Nuvilight 1000 250 650 0.38 for illumination
[Luc05] 2005 Luceat SI-Type 1000 150 650 0.46 30 MHz100 m
[Wal05] 2005 Nanoptics A-POF 1000 100 650 - conception
[Hai05] 2005 Huiyuan SI-POF 1000 300 650 - coextrusion
[Zie06h] 2006 Luceat SI-POF 1000 135 650 0.50 from preform
65 520
5,000
attenuation [dB/km]
2,000
1,000
500
200
100
wavelength [nm]
50
400 450 500 550 600 650 700 750 800
Fig. 2.59: Attenuation of different standard-NA SI-POF (measurement by POF-AC)

It was not until about 1980 that technology made possible the production of
POF which came relatively close to the theoretical attenuation minima. Initial
problems with the service life and with certain mechanical loads were quickly
solved with on-going developments. In Fig. 2.59 the spectral attenuation curves of
three SI-POFs are shown (data sheet information). All three fibers from Japanese
manufacturers are close together. The visible differences may possibly be due to
different methods of measurement.
Most manufacturers offer SI-POFs in different diameters. In [Zub01b] and
[Nuv04] the properties of these fibers are compared (Table 2.5).
Table 2.5: Attenuation of POF with different diameter

Attenuation [dB/km]
diameter [μm] 250 500 750 1.000
Mitsubishi < 700 < 190 < 180 < 160
Toray < 300 < 180 < 150 < 150
Asahi Chem. n. a. < 180 < 180 < 125
BOF < 150 < 150 < 150 < 150
Optectron < 150 < 150 < 150 < 150
Nuvitech < 350 < 250 < 250 < 250
For Toray fibers, the losses of fibers with different diameters are listed in the
data sheet and are shown in Fig. 2.60.
10000
attenuation [dB/km]
3000
1000
300 core t 750 μm

core = 500 μm
100 core = 250 μm
wavelength [nm]
30
400 500 600 700 800 900
Fig. 2.60: Attenuation of different PMMA-SI-POF by Toray
With a few exceptions the losses for all fiber diameters are similar. Some
reasons for the increase in attenuation with thinner fibers could be that either the
high attenuation of the optical cladding plays a greater role or that more stress is
exerted on the thin fiber during manufacture. A fiber with a ¼ mm core diameter
has only one sixth the thermal capacitance. When the cladding and opaque jacket
are applied this fiber is necessarily warmer. The process temperatures during ma-
nufacture can indeed lie clearly above the glass transition temperature.
The youngest manufacturer of PMMA SI-POF is the Italian company Luceat.
Here fibers for diverse applications, mainly in mechanical engineering, are pro-
duced. The highest quality is still in the developmental stage. A comparison of the
measured values of Luceat fibers (POF-AC 2006, [Ziem06h]) with the values
from [Wei98], more or less the POF reference curve up until now, is shown in
Fig. 2.61.
500
attenuation [dB/km]
300
200
Luceat
100
80 [Wei98]
60
50
400 450 500 550 600 650 700
wavelength [nm]
Fig. 2.61: Attenuation of SI-POF by Luceat (2006)
In the area of 520 nm this fiber is even somewhat better that the data of the best
fibers so far. Thanks to the availability of reasonably priced and fast green LEDs
this advantage can be assessed very highly. As part of the European POF project
POF-ALL (see www.ist-pof-all.org) the transmission of a 10 Mbit/s data stream
was able to be demonstrated over 425 m (see System Chapter).
2.3.3 Double Step Index Profile Polymer Fibers
We have already discussed the principle idea of a double step index profile POF.
All three important Japanese manufacturers presented such fiber types around
1995. After the expectations that ATM would become the dominating network
technology in the home were not fulfilled, these fibers have more or less become
niche products today, albeit at relatively high prices. Today in many areas there is
a demand for data rates which require the use of these fibers instead of the normal
SI-POFs. Technically, DSI-POFs are on a comparable level and would hardly be
more expensive than SI-POFs when produced in high volumes.
Table 2.6 compares the properties of DSI-POFs of the three manufacturers

([Mit01], [Nich03], [LC00b]).
Table 2.6: Overview of DSI-POF

Mitsubishi Toray Asahi
MH4001 PMU-CD1001 AC1000(I)
diameter [μm] 980 1000 ± 45 1000 ± 60
attenuation (650 nm) [dB/km] 160 170 160
numerical aperture - 0.30 0.32 0.25
bandwidth MHz km 10 >10 15
temperature range [°C] -55 .. +75 -20 .. +70 -40 .. +70
bend radius [mm] 25 - 25
We would like to point out once again that the DSI-POFs are usually offered
now as before as low NA POF. In the first few years manufacturers did not pro-
vide any information at all about the double cladding structure. In [Eng98b] the
double cladding structure was proven quite early on the basis of measurements of
the far field and with optical microscopy. In Fig. 2.62 you can see the far field dis-
tributions for different fiber lengths measured with the inverse far field method at
the FH Gießen/Friedberg.
1.0
Popt
1m
0.8 10 m
50 m
0.6 O= 594 nm 90 m
0.4
0.2
0.0 4 [°]
-30 -20 -10 0 10 20 30
Fig. 2.62: Inverse far field measurement of a DSI-POF
You can clearly see that after short distances much light from the interface
layer between inner and outer cladding is still guided. After 50 m these shares
have disappeared and the angle distribution corresponds to a true low NA POF.
Figure 2.63 shows two microscope photos of DSI-POF (Univ. of Ulm). Both
optical claddings can be easily recognized.
At the 2003 POF Conference Mitsubishi was the first manufacturer to present
the actual structure. The effect of suppressing higher modes by high attenuation of
the inner cladding was also confirmed theoretically and experimentally. For
example, Asahi gives a value of 6000 dB/km at 650 nm for the losses in the inner
cladding.
Fig. 2.63: Double cladding structure of a POF
2.3.4 Multi-Core Polymer Fibers
Since 1994, polymer fibers as multi-core fibers have been introduced, e.g. in
[Tesh98], [Mun94], [Asa97] and [Tesh98]. Table 2.7 shows a few parameters
from these publications.
Table 2.7: Multi-core POF (Asahi Chemical)
Type Ref. No. of Structure NA Attenuation Bandwidth

Cores at 650 nm
NMC-1000 POF´94 19 SI 0.25 125 dB/km 170 MHz100 m
PMC-1000 Data´96 217 SI 0.15 270 dB/km n. a.
MCS-1000 Data´97 217 SI - 320 dB/km n. a.
- POF´98 37 DSI 0.19 155 dB/km 700 MHz50 m
- POF´98 37 DSI 0.25 160 dB/km n. a.
- POF´98 37 DSI 0.33 160 dB/km n. a.
NMC-1000 Data´98 37 DSI 0.25 160 dB/km 500 MHz50 m
PMC-1000 Data´98 37 DSI 0.19 160 dB/km n. a.
- POF´98 217 SI 0.50 160 dB/km n. a.
- POF´98 217 SI 0.33 160 dB/km n. a.
The MC-POF features a noticeably reduced sensitivity to bending and only in-
significantly increased attenuation as well as a significantly increased bandwidth
compared to single core fibers, this being due to the possibility of smaller nume-
rical apertures. Whether these fibers can be produced at the same price is still an
open question. Should this be possible, data rates of 500 Mbit/s up to 1 Gbit/s over
50 m can easily be achieved in commercial applications. At the POF-AC a data
rate of over 1 Gbit/s over 100 m MC-POF has already been achieved.
At present, only Asahi chemical offers MC-POF for data communication while
other manufacturers offer this kind of fiber for lighting purposes or also as image
guiding fiber. The following photos show the cross-sections of the three, presently
available MC-POFs with 37, 217 and 631 cores (the 19 core variant is no longer
available).
Fig. 2.64: Photo by microscope of MC-POF, 37, 217 respectively 631 cores
An overview of the technical data of the four different MC-POFs is summa-

rized in the following Table 2.8 (Nichimen data sheets). The PMC 1000 permits
the highest data rates on the basis of experiments conducted so far since it posses-
ses a DSI structure.
Table 2.8: Data of MC-POF
Parameter Unit MCQ-1000 MCS-1000 NMC-1000 PMC-1000

number of cores - 613 217 19 37
single core μm 37 60 200 130
core material - PMMA PMMA PMMA PMMA
cladding material - fluoro polymer FMA-copolymer
2nd cladding material - - VDF-copolymer
NA - 0.5 r 0.05 0.50 0.25 0.19
fiber core mm 1.0 r 0.06 1.0 r 0.06 1.0 r 0.06 1.0 r 0.06
cable mm 2.2 r 0.07 2.2 r 0.10 2.2 r 0.10 2.2 r 0.10
jacket PE PE (black) PE (black) PE (black)
attenuation1) dB/km <200 320 1634) 1634)
5)
bit rate (50 m) Mbit/s n. a. n. a. 350 5005)
temperature °C -40 .. +60 -40 .. +60 -40 .. +70 -40 .. +70
bend loss2) dB <0.1 <0.23) <0.13) <0.13)
1) 4)
Cut back 12 - 2 m, at 650 nm 650 nm monochromatic light
2) 5)
R = 3 mm, 180°, no stress 650 nm LD, BER = 10-12
3)
R = 3 mm, 360°, Launch-NA: 0.2
We do have to point out one special feature of these four MC POFs: the fibers
are tightly bound in the cable as opposed to the individual fibers in a fiber glass
bundle or other MC POFs used in lighting technology. The share of the core
surface is not only enlarged, but it is considerably easier to work the fibers. These
strands can be mounted like quite normal 1 mm SI-POFs.
The two enormous advantages of MC-POF, namely the high band width and
the low bending losses, have in the meantime been somewhat qualified since con-
siderably cheaper GI-POFs on a PMMA basis have become available. The latter
will be treated in the next paragraph.
2.3.5 Multi-Step Index Profile and Graded Index Profile Fibers
The greatest bandwidths of all fibers - with the exception of the singlemode fibers
- are shown by graded index profile fibers. They have been used extensively for
some time in the field of silica glass fibers and are a standard. In the USA, predo-
minately fibers with a core diameter of 62.5 μm are used, whereas in Europe and
most other countries fibers with a core diameter of 50 μm are used. This diameter
is nevertheless 5 to 6 times greater than with singlemode fibers whereby the plug
costs are greatly reduced and the coupling of lasers is also easier. The bandwidth-
length product (BLP) of these multimode glass fibers lies in the range of 200 to
500 MHz · km. For the transmission of 10 Gbit/s a new fiber specification with a
BLP of 2,000 MHz · km at a wavelength of 850 nm is even being developed (for
example, see [Oeh02] and [Geo01]).
The advantages of the large core diameter and high bandwidth would be an
optimal combination with POFs. Furthermore, numerous problems with the core-
cladding interface area would cease to exist with GI fibers since the light guiding
would take place exclusively in the core. Glass GI fibers are produced by applying
many layers of a SiO2-GeO2 mixture with different compositions to a quartz glass
pipe. Finally, the fiber is drawn (several 100 km) out of such a preform. Unfor-
tunately, this is not possible with POFs. The different methods and combinations
of materials with which attempts have been made to produce GI-POF will be des-
cribed further on. Since GI fibers are difficult to produce - as we shall describe
later on - a series of multi step index profile POFs have been introduced. These
MSI-POFs also offer high bandwidth depending on the number of steps. For now,
the optical characteristics are summarized here.
Table 2.9 shows an overview of the values for PMMA-based GI, MC and MSI
fibers. To the best knowledge of the author, all PMMA-GI-POF published to date
are produced by doping, whereas only MSI-POF are produced in a co-polymeri-
zation process.
Table 2.9: Published data of PMMA-GI-, MSI- and MC-POF (IGPT: interfacial gel poly-
merization technique; PFM: preform method)
Ref. Year Producer Material Øcore Attenuation at O NA Remarks
μm dB/km nm
[Koe98] 1998 1 11 8 100 m
[Koi95] 1982 Keio Univ. MMAco VPAc - 1070 670 - first GI-POF
[Koi96c] 1990 Keio Univ. PMMA - - - - 670 nm: 300 MHzkm
[Koi95] 1990 Keio Univ. MMA co VB - 130 650 -
[Koi90] 1990 Keio Univ. MMA-VB - 134 652 - IGPT, 260 MHz1 km
[Koi90] 1990 Keio Univ. MMA-VPAc - 143 652 - IGPT,125 MHz 1km
[Koi92] 1992 Keio Univ. PMMA 200-1500 113 650 - IGPT, 1,000 MHzkm
[Koi92] 1992 Keio Univ. PMMA 200-1500 90 570 -
[Non94] 1994 Sumitomo PMMA 400 160 650 0.26 'n=0.014, 8GHz50m
[Shi95] 1995 BOF PMMA 600 300 650 0.19 3 GHz100 m
Table 2.9: Published data of PMMA-GI-, MSI- and MC-POF, continued

Ref. Year Producer Material Øcore Attenuation at O NA Remarks
μm dB/km nm
[Ish95] 1995 Keio Univ. PMMA-DPS 500-1000 150 650 - 585 MHzkm
[Koi97b] 1997 Keio Univ. PMMA - - - - 2 GHz100 m
[Tak98] 1998 Kurabe PMMA 500 132 650 - 2 GHz100m, PFM
[Tak98] 1998 Kurabe PMMA 500 145 650 - 2 GHz90m, PFM
[Tak98] 1998 Kurabe PMMA 500 159 650 - 680 MHz50m, PFM
[Tak98] 1998 Kurabe PMMA 500 329 650 - PFM
[Mye02] 2002 Dig. Optr. Polymer 180 350 685 0.20 no samples available
[Shin02] 2002 KIST Korea PMMA 1000 120 650 0.26 g=2.4; 3.45 GHz100m
[Liu02a] 2002 Huiyuan PMMA - - - - since 2001
[Luv03] 2003 Luvantix PMMA ? 160 650 0.33 3.5 GHz bandwidth
[Fuj04] 2004 Lumistar PMMA 500 3 Gbps50m
[Rich04] 2004 Optimedia PMMA 900 200 650 0.40 commercially available
[Yoo04] 2004 Optimedia PMMA 675 200 650 0.40 commercially available
[Nuv05] 2005 Nuvitech PMMA 500 180 650 0.25 3 Gbps50m
[Nuv05] 2005 Nuvitech PMMA 900 180 650 0.30 3 Gbps50m
[Fuj06] 2004 Lumistar-X new low loss 120 100 850 ? 10 GHz50m
MSI-POF
[Shi99] 1997 Mitsubishi PMMA 700 210 650 0.30 500 MHz50m, 4-7 layers (?)
(Eska-Miu)
[Lev99] 1999 RPC Tver PMMA/ 4FFA 800 400 650 7 layers, 310 MHz100 m
From the beginning of the 90s, it became possible to produce PMMA-GI-POF

having an attenuation at 650 nm that is similar in quality to that of SI-POF. In
doing so, it was possible to attain bandwidths up to 50 times larger which are ade-
quate for transmitting several Gbit/s across distances up to 200 m. Likewise,
multi-core and multi-step index POF achieve similar values for attenuation and
allow data rates up to 1 Gbit/s across distances of 50 m, e.g. for applications in
compliance with IEEE1394 (up to S800). The core diameter of all these fibers
typically lies between 0.5 mm and 1 mm which means that existing reasonably
priced connectors can be used.
Multi-step index profile fibers - the last lines in the table - have been described
in [She99] and [Lev99]. In the group headed by Prof. Levin, different materials
were used for the production of layers with different refractive indices
(P(MMA/4FFA), P(MMA/4FMA) and PMMA-naphthalene). The best results
were obtained with the mixture PMMA/4FFA which has an attenuation of
approximately 400 dB/km (at 650 nm) and a bandwidth of 310 MHz·100 m. The
total number of 7 steps in the fiber with a core diameter of approximately 800 μm
were produced in a preform and subsequently drawn.
The ESKA-MIU has a core/cladding diameter of 700 μm/750 μm and also has
several layers (probably between 4 and 7) which are produced by co-polymeri-
zation. Originally, 4 to 7 layers were presumably being aimed at but in the end this
fiber with 3 layers was achieved as a product. It is said to be produced in a con-

tinuous drawing process. The bandwidth in [Shi99] is stated to be larger than
500 MHz 50 m. In several publications this fiber is called a GI-POF ([Sak98],
[Num99]). The difference between this design and “genuine” GI fibers is prima-
rily the larger core diameter. In [Num99] the attenuation of the fiber is stated as
being 210 dB/km with an AN = 0.30, i.e. values that are comparable with the DSI-
POF. Materials and measurements of the index profile will be discussed in the
Section Production and Materials.
Institutes and companies from South Korea have been very successful in
producing PMMA GI-POF. In the past few years publications have come from:
¾ Department of Materials Science and Engineering, Kwangju Institute of
Science and Technology (KIST), Kwangju
¾ Center for Advanced Functional Polymer, Department of Chemical Enginee-
ring, KAIST, Taejon, Korea
¾ E-Polymer Laboratory, SAIT, Taejon, Korea
¾ Optics Laboratory, Seoul, Korea
¾ Optimedia, Korea
¾ Nuvitech, Korea
¾ Luvantix, Korea
The production method for GI POF is described in [Shin03]. A MMA-BzMA
mixture is poured into a rotating cylinder. The purpose of the rotation is simply to
form even concentric layers. The polymerization takes place thermally and the
concentration of BzMA is continuously increased to 15%. This emerging preform
is then drawn into a fiber. Figure 2.65 shows the pulse broadening for a 66 m long
fiber which corresponds to a BLP of 3.45 GHz · 100 m. The smallest measured
attenuation of the fiber is given at 120.6 dB/km.
1.2
Intensity [a.u.]
1.0
1m 66 m
0.8
0.6 99.5 ps 129.5 ps
0.4
0.2
0.0
t [ps]
-0.2
0 200 400 600 800 1000 1200 1400 1600
Fig. 2.65: Pulse broadening in PMMA-GI-POF ([Shin03])

Since 2004, a new GI-POF on a PMMA basis has been available on the market.
The OM-Giga (see [Rich04] and [Yoo04]) has a core diameter of 900 μm or
675 μm respectively and a nearly parabolic profile. It is produced through poly-
merization of several layers, although the steps are almost completely smoothed
through thermal treatment. According to the data sheets available in the Internet
the fibers have the following parameters (Table 2.10).
Table 2.10: Parameter of GI-POF OM-Giga

Property Unit B-075 B-100 Remarks
core diameter μm 750 (675) 1,000 (900) (GI-region)
diameter variations % ±5 ±5
tensile strength N > 35 > 65 at break
bend radius mm 25 25
temperature range °C -30 .. +60 -30 .. +60
attenuation dB/km < 200 < 200 at 650 nm
bandwidth GHz > 1.5 > 1.5 for 100 m
The fact that this fiber possesses thermal stability comparable to a standard
POF, different from GI-POF with doping, must be rated as a particularly great
step. Even after 5,000 hours of operation at 80°C no change in the bandwidth
could be determined. The cross-section of a 1 mm OM-Giga is shown in Fig. 2.66
(microscope photograph shown in wrong colors). The approx. 10 index steps can
still be seen quite well.
Fig. 2.66: Cross section of an OM-Giga (POF-AC) and a MSI (Tver, [Ald05])
In Fig. 2.67 the change in the refractive index profile of a doped PMMA
GI-POF is shown after accelerated aging (122 hours at +109°C, from [Bly98a] and
[Bly98b]). You can see quite well that the index profile is still parabolic at the
beginning of the aging process. The share of the dopants is the greatest in the
center of the fiber which is why the glass transition temperature has sunk the most.
The dopant diffuses outwardly. Consequently, the concentration increases out-

wardly, Tg also drops there and the diffusion process continues until the profile has
almost become rectangular. The attenuation of the fiber will hardly increase, but
the bandwidth drops dramatically. Decisive for the application temperature of the
fiber is the dopant concentration in the axis.
Fig. 2.67: Change of the refractive index profile of a GI-POF by ageing
Measurements on OM-Giga at the POF-AC will subsequently be introduced.

The results of a long-term temperature test are shown in Fig. 2.68. The bandwidth
was measured for over 5,000 hours on a 50 m long sample in order to be able to
determine changes in the index profile.
measured bandwidth [MHz]

2500
2000
1500
1000
500 70°C 80°C
time [h]
0
0 400 800 1200 1600 2000 2400
Fig. 2.68: Long-term behavior of OM-Giga
The frequency range of the network analyzer extended to 1.3 GHz. The values
represented were determined through extrapolation and thus burdened with a rela-
tively large error. A clear deviation from the parabolic index profile would in any
event have caused a very strong decrease in the bandwidth.
The stable bandwidth proves that co-polymerization is obviously a suitable
means to produce thermally stable and thus long-life PMMA GI-POF.
A comparison of the measured attenuation of ESKA-MIU and OM-Giga is
shown in Fig. 2.69. The attenuation of the OM Giga is somewhat higher at 650 nm
than that of the Mitsubishi fiber and also of the SI-POF. However, it clearly shows
the greatest bandwidth.
1000
attenuation [dB/km]
800
600
OM-Giga
400
217 dB/km
300
200
ESKA-MIU 161 dB/km
100
400 450 500 550 600 650 700
wavelength [nm]
Fig. 2.69: Spectral attenuation of ESKA-MIU and OM-Giga
The Korean manufacturer Luvantix offers preforms for PMMA GI-POF

([Luv03] and [Kim03]). The index profiles from both refernces - measured values
and approximation of each - are shown in Fig. 2.70.
The authors do not know what relations exist between Luvantix as preform
manufacturer and Nuvitech and Optimedia as fiber producers as well the different
research institutes. Overall, however, POF production in South Korea seems to
enjoy greater attention and more progress in the field is foreseeable.
Other announcements concerning the production of GI-POF came from the
USA (Digital Optronics and Nanoptics, [Wal02], [Mye02] and from China
[Liu02a]). Since no data or even fibers are known from these producers they will
not be considered in greater detail.
In Sections 2.5 and 2.6 data on bending behavior and bandwidth are summa-
rized. The production methods are presented in Section 2.8.
2.4 Glass Fibers for Short-Range Data Transmission 93
refractive index refractive index

1.512 1.510
AN = 0.21 AN = 0.32
1.510 1.505
1.508 1.500
1.506 1.495
1.504 1.490
1.502 1.485
1.500 1.480
theory theory
1.498 1.475
measured measured
1.496 1.470
1.494 1.465
rel. radius (r/rc) rel. radius (r/rc)
1.492 1.460
0.0 0.2 0.4 0.6 0.8 1.0 0.2 0.4 0.6 0.8 1.0
Fig. 2.70: PMMA GI-POF index profile (left: [Kim03], right: [Luv03])
2.4 Glass Fibers for Short-Range Data Transmission
2.4.1 200 μm Glass Fibers with Polymer Cladding
Thanks to their simple production and great robustness silica glass fibers with
polymer cladding have been used for a long time. Figure 2.71 shows the principle
structure. A core (typically with a diameter of 200 μm) of homogeneous SiO2 is
surrounded by a high-strength, transparent polymer with smaller refractive indices
(about 15 μm thick).
500 μm
inner jacket
2.3 mm
outer jacket 200 μm
SiO2-core
230 μm
polymer
Fig. 2.71: Structure of a 200 μm PCS
Production is so easy because the core is drawn from a quartz glass cylinder.
The polymer cladding is applied by extrusion after it has cooled off. First of all, all
glass fibers are extremely sensitive to water and must be protected by a plastic
coating as thick as possible. Furthermore, pure glass fibers do not have a great
94 2.4 Glass Fibers for Short-Range Data Transmission
mechanical load capability. The polymer cladding gives the fibers the capacity to
bear extreme loads. The jacketed fiber can thus hardly be shattered. Pure glass-
glass fibers (glass core with an optical glass cladding) are always surrounded by
similar protective layers, e.g. acrylates which, however, do not have any optical
function.
Because of its refractive index and attenuation the polymer cladding determines
to a great extent the optical parameters of the PCS. In short wavelength ranges the
attenuation nearly corresponds to pure SiO2 fibers. Above approx. 1,000 nm the
losses in the polymers are so high that the effective PCS attenuation also rises ra-
pidly. Silica glass can endure temperatures up to 1,000°C, but not the polymer
cladding. Consequently, the primary coating material determines the thermal and
chemical characteristics. Most PCSs available in the market have been specified
for an application temperature of +70°C. Some more recent types have been di-
mensioned for use in automobile networks for temperatures up to +125°C. Infor-
mation on such PCSs can be found for example in [Hub03] and [Schö03]. Fig 2.72
has been taken from the latter work. You can clearly recognize how strongly the
attenuation spectra of different PCSs can depend on the cladding materials
selected.
10,000
attenuation [dB/km]
1000
100
diff. PCS
10
theoretical limit
0.1
200 400 600 800 1,000 1,200 1,400 1,600 1,800
wavelength [nm]
Fig. 2.72: Attenuation of different 200 μm PCS according to [Schö03]
Just as with glass-glass fibers the absence of water plays an important role for
PCS for keeping losses low especially in the long-wave ranges. So-called all-silica
fibers in which the optical cladding consists of silica glass are used at high tem-
peratures. These fibers are also employed for the transmission of very high light
power (working with lasers) since it is very important that no light is absorbed at
the core-cladding interface layer.
Table 2.11 lists some of the representative types taken from a number of diffe-
rent PCS variants which differ in cladding material, core diameter and NA (data
from [Hub03] and [OFS02]).
Table 2.11: Properties of different PCS

Parameter Unit All Silica All Silica HCS HCS PCS
High OH Low OH High NA Low OH [Hub03]
producer OFS OFS OFS OFS Polymicro
core/ μm 200/240 200/240 200/230 125/140 200/230
cladding 365/400 365/400 400/430 200/230
550/600 550/600 300/330
940/1000 940/1000 400/430
NA - 0.22 0.22 0.43 0.37 0.37
D (820 nm) dB/km 10 8 6 12 6
10 8 8 6
10 8 8
10 10 8
bandwidth MHz km n. a. n. a. n. a. 20 20
20
15
13
bend radius mm 14 14 16 15 16
(long term) 47 47 47 16
94 94 24
118 118 47
temperature °C -65..+135 -65..+135 -65..+125 -65..+125 -40..+125
PCS are generally used in lengths of up to a maximum of 200 m. The attenu-

ation then only amounts to a few dB which can for the most part normally be dis-
regarded. If a LED is used as a transmitter, considerably less light will be coupled
into the fiber than into a 1 mm POF. On the other hand, the light can be coupled
more effectively into the photodiode. Different manufacturers even offer trans-
mission systems which can work with the same plug construction with POF as
well as with 200 μm PCS, e.g. [HP01].
system with
1 mm POF
system with
Popt [dBm] 200 μm PCS
-4 -6 -8 -10 -12 -14 -16 -18 -20 -22 -24 -26 -28 -30
LED-power range (launched into the fiber)
receiver sensitivity range
allowed path loss (with margin)
Fig. 2.73: Link power budget for POF and PCS
Fig. 2.73 shows power budgets for both possibilities, each with the same trans-
mitters and receivers (system for 125 Mbit/s).
The result for PCS is a permissible fiber attenuation of at least 11 dB - taking
the system margin into consideration - thanks to the greater input power. In this
way at least 20 m of POF can be bridged. The guaranteed loss for PCSs is only
7 dB. However, at least 100 m of fiber can be bridged, limited here due to the
bandwidth.
max. bit rate [Mbit/s] max. bit rate [Mbit/s]

200
max. at +25°C max. at +25°C
100
70
40 recommended recommended
application application
area with POF area with PCS
20
fiber length [m] fiber length [m]

10
10 20 30 40 60 80 100 10 20 50 100 200 500 1000
Fig. 2.74: System parameters of the HP-system with POF and PCS (according to [HP01])
The bandwidth for PCS indicated in the data sheets has to be viewed with a
certain degree of skepticism. Measurements conducted at the POF-AC show that
all PCSs investigated with An = 0.37 at full launch have a BLP in the range of
5-7 MHz · km. This lies clearly below the specified data of 10-20 MHz · km. This
is not a contradiction, however, since none of the manufacturers as a precaution
provided any information about the measurement conditions. One reason may be
that the PCS was developed for relatively low data rates (10 Mbit/s and less). The
fiber bandwidth therefore did not play any role whatsoever while the POF was
also designed from the very beginning for higher data rates. Diverse information
and publications on bandwidth exist for the different polymer fibers as summa-
rized in Chapter 2.5. The most recent draft for the standardization of PCS is
viewed by the IEC as having a bandwidth of 5 MHz · km for fibers with a NA of
0.40 ± 0.04.
A specific problem with PCS in the past was that the temperature coefficients
of glass and plastic did indeed deviate considerably from one another. In the case
of some fibers this resulted in a refractive index difference - and NA, too - which
dropped to zero at low temperatures. This effect is shown in Fig. 2.75 taken from
[Dug88].
Far field distributions are represented in the picture after 2 m of fiber at diffe-
rent temperatures. They were measured with laser stimulation at altered angles. In
this case the optical cladding was a silicone plastic. Modern PCSs no longer show
this effect.
rel. power
1.0
0.9 +40°C
0.8 -2°C
-31°C
0.7 -51°C
0.6 -65°C
-72°C
0.5 -92°C
0.4 -98°C
0.3
0.2
0.1
T [°]
0.0
-5 0 5 10 15 20 25 30 35
Fig. 2.75: Temperature dependence of PCS-NA, presented as far field
2.4.2 Semi-Graded Index Glass Fibers
Up until some time ago this class of fibers was only available as a product from
the manufacturer Sumitomo ([Sum03]). Except for the gradients introduced this
fiber corresponds to conventional PCS. The index variation is attained by adding
germanium which is also usual for silica glass. Even with normal 50 μm GI fibers
the germanium share represents a considerable cost factor. The semi-GI PCS,
however, has a 16-fold cross-section. This type of fiber is still extremely expen-
sive. It is still open how far the price can drop when manufacturing greater
lengths. In the meantime, OFS has appeared as a second manufacturer
([Ziem06i]).
100
80 spectral attenuation [dB/km]
60
40 Sumitomo
30
20 OFS
10
8
6
4 wavelength [nm]
450 500 550 600 650 700 750 800 850 900 950 1000
Fig. 2.76: Spectral attenuation of the semi-GI-PCS
Figures 2.76 and 2.77 show the attenuation curve and the pulse response of the
semi-GI-PCS based on measurements made at the POF-AC. The following table
gives the parameters from the data sheet - the bending radius and the operating
temperature are not specified. The bandwidth and maximum data rate measure-
ments are dealt with in the corresponding sections.
opt. power [a.U.]

1.0
full mode launch
500 m PCS
0.8
0.6 FWHM: 6.8 ns

| 32 MHzkm
0.4
0.2
t [ns]
0.0
0 10 20 30
Fig. 2.77: Pulse response of Semi-GI-PCS
Table 2.12: Parameters of Semi-GI-POF

Parameter Unit HG-series Semi-GI V2
Sumitomo OFS
core μm 200 200
cladding μm 230 230
core structure n. a. n. a. VAD/MCVD GI
NA - 0.40 0.36
GI-NA n. a. 0.275
D (820 nm) dB/km 6 8
bandwidth MHz km 100 48 (overfilled)
2.4.3 Glass Fiber Bundles
2.4.3.1 Quartz Glass Fiber Bundles

Glass fiber bundles are employed in the most diverse areas. Above all, it makes
sense to use them when a large light-guiding cross-section is to be combined with
high cable flexibility. In optical measurement techniques bundles of quartz glass
fibers are employed which permit a continuous high transmission rate in the range
between 380 nm and 2000 nm. If you lay out the fibers differently at both ends of
the cable, then they can also serve as cross-section converters, e.g. monochro-
mators. The end surfaces are usually prepared: the bundle is glued in the plug and
then polished. Figure 2.78 shows an example.
Fig. 2.78: Example for a quartz glass fiber bundle
The transmission of such a bundle is shown in Fig. 2.79 (acc. to [Ori01]). The
greatest part of the 100% missing share is determined by the only about 60% part
of the core surfaces and the Fresnel losses. The numerical aperture of the bundle
shown is 0.22, the length is about 1 m, and the single fiber diameter is 200 μm.
70%
transmission
60%
50%
40%
30%
20%
10%
wavelength [nm]
0%
200 400 600 800 1,000 1,200 1,400 1,600 1,800 2,000
Fig. 2.79: Transmission of a quartz glass fiber bundle [Ori01]

2.4.3.2 Glass Fiber Bundles

Pure quartz glass is many times more expensive than normal polymers but also as
conventional mineral optical glasses. Attenuations of some 100 dB/km is absolu-
tely acceptable for many applications, e.g. in lighting technology. Schott has been
producing bundles of thin glass-glass fibers for quite some time. The index diffe-
rence can be varied in diverse areas by choosing a particular glass composition.
The spectral attenuation of a typical glass fiber bundle compared with a PMMA
POF is shown in Fig. 2.80. The glass has higher losses in the blue areas whereby
the cable length is limited when guiding white light. The POF has greater losses in
the near infrared range.
1,000
D[dB/km]
500 POF
200 MC-GOF
100
O [nm]
50
500 550 500 550 600 650 700 750 800 850
Fig. 2.80: Spectral attenuation of glass fiber bundles and POF
An entirely new application for such glass fiber bundles has come about with
the ever increasing use of optical networks in vehicles. The previous systems are
specified with 1 mm POF. Two parameters especially limit the use: the tempera-
ture range is limited to a maximum of +85º and the relatively large bending radius.
Both limitations can be reduced considerably with glass fiber bundles, whereby
the usual optical characteristics are for the most part retained so that the identical
active components can be used. Table 2.13 from [Lub04b] compares the parame-
ters of a glass fiber bundle (MC-GOF) with those of a POF for vehicle networks.
The construction of the plug is especially problematical. The usual method of
cementing and polishing takes too much time for mass production and results in
the core surface having too low a share with correspondingly high losses with the
plug connections.
Megomat TS AG, working together with Schott, has developed a new kind of
assembly procedure ([War03]). The actual fiber bundle has a diameter of 1.2 mm.
The plug has a metal ferrule with a corresponding opening. During production the
fiber bundle is heated to such an temperature that the glass can be compressed.
The fibers are pressed closely together when crimped so that the diameter of the
bundle is lowered to 1 mm.
Table 2.13: Comparison MC-GOF and POF
Parameter Unit MC-GOF MOST-POF

core diameter [μm] 53 1,000 r 45
cladding thickness [μm] 3 10
number of cores - approx. 400 1
ncore/ncladding - 1.585 / 1.49 1.49 / 1.40
numerical aperture - 0.50 0.50
attenuation at 650 nm [dB/km] 250 160
bandwidth (full launch) [MHz·20 m] 150 >50 (200 typ.)
bend radius [mm] 5 25
temperature [°C] -40 .. 125 -40 ..85
The core share of the plug end face then amounts to about 85%. After the
crimping the bundle is broken off and polished. Figure 2.81 shows a photo of the
plug end face.
Fig. 2.81: Photograph of a MC-GOF
The irregular arrangement of fibers in the bundle leads to different patterns

when pressed together. Neighboring fibers mostly form regular hexagonal struc-
tures which can, however, also have big gaps. Every once in a while linear struc-
tures with pentagonal fibers are formed. The individual fibers are deformed at the
edge of the bundle, something which happens quite irregularly.
Fig. 2.82: Details of MC-GOF connector end faces
Fig. 2.83: Details of MC-GOF connector end faces
Since the bundle consists of about 400 individual fibers this irregular defor-
mation of individual fibers does not play any role overall. Since the deformations
only arise over a few millimeters there is no significant additional attenuation.
In conclusion, Fig. 2.84 shows an x-ray photo of the bundle within the cable.
The individual fibers have to move freely within the cable. When there is a tight
bend the change in length is distributed on the inside and outside for a long stretch
so that the fibers are only subject to a slight load. That is why the bundle can take
bending radii of only a few mm.
Fig. 2.84: X-ray picture of a bundle

2.5 Bandwidth of Optical Fibers 103
2.5 Bandwidth of Optical Fibers
In order to be able to determine the bandwidth of an optical fiber, several different

influencing factors need to be considered. Mode dispersion and chromatic disper-
sion are the most important factors involved in multimode fibers. Particularly in
the case of POF, mode dispersion depends on various parameters such as wave-
length, the light launching conditions, refractive index profile, fiber-laying condi-
tions as well as the homogeneity of the fiber's characteristics. In the following
sections we intend to show how the conclusions drawn from basic physical pro-
cesses can be used to explain values that are measured in reality.
2.5.1 Definition of Bandwidth
It is possible to define the term bandwidth in quite different ways. Essentially it

describes the frequency range of a system within which the transmission of signals
can be achieved with reasonable attenuation. In POF systems, the limiting factor is
usually the bandwidth of the fiber itself, which is created by modal dispersion. As
we will demonstrate later on, one can describe the SI-POF as being very close to a
Gaussian low pass filter. In this book we will use the following definition of band-
width:
f3dB : Frequency at which the amplitude of a sinus modulated monochromatic
signal has been reduced to ½ of the optical level (see Fig. 2.28).
Figure 2.85 schematically illustrates this definition.
rel. opt. amplitude at

fiber output
1.0 frequency response
0.5
bandwidth f3dB
frequency f
0.0
Fig. 2.85: Definition of POF bandwidth
Nonetheless, knowledge of the bandwidth alone is not adequate for estimating

what the actual capacity of the complete link will be. In order to determine this
parameter, it is further necessary to know the actual transmission procedure as
well as the complete transmission function. For example, it is possible to transmit
signals of significantly broader bandwidth if an electrical compensation of the
frequency response takes place, as illustrated schematically in Fig. 2.86.
104 2.5 Bandwidth of Optical Fibers
P(f)POF P(f)compensating filter P(f)resulting

1.0
0.5
f [a.U.] f f
0.0
0 1 2 0 1 2 0 1 2
Fig. 2.86: Compensation of the POF low pass characteristic
A high pass filter is used for compensation. In the case of low frequencies, the
signal is attenuated - in the case of higher frequencies the signal is passed through
without attenuation. The resulting function has a significantly higher bandwidth;
however, due to the overall existing level of attenuation, a higher level of signal is
necessary.
In addition, the type of signal involved (digital or analogue) is also of signifi-
cance and finally the required system reserves must be considered. The following
general relationship can be used as a rule of thumb for digital systems:
maximum bit rate [Mbit/s] = 2 u bandwidth [MHz].
We intend here to look at bandwidth as a function of fiber characteristics. For
this reason, the effect of chromatic dispersion will be initially neglected because it
is directly proportionally dependent on the spectral width of the source.
In this section we will show experimental investigations on the bandwidth of
SI-POF fibers. After explaining the measurement procedures, we will show to
what extent bandwidth is particularly dependent on the launching conditions.
2.5.2 Experimental Determination of Bandwidth
As a frequency response multimode fibers show an almost Gaussian-like behavior:
P( f ) P0 e
f 2 / f02
As can be easily demonstrated, the amplitude of a Gaussian low pass filter
(P(f) = P0 · exp (f²/f0²)) for f = 1.17741 · f0 has dropped to half the value that
applies for f = 0. When using a spectrum analyzer to measure the frequency
response of a fiber link, it is necessary to determine the electrical 6 dB width
because the photodiode will convert the optical power proportionally into a
current. Therefore the following applies:
2
Pel Popt
Figure 2.87 shows an example for such a bandwidth measurement for a 30 m

standard NA-POF.
-3
electr. power [dBm]
measured
-6
15 m SI
-9 650 nm
239 MHz
NA: 0.34
-12
-15 Gaußian
approx.
-18
1 10 100 f [MHz] 1,000
Fig. 2.87: Bandwidth measurement at a SI-POF
Due to the limited dynamics of the measurement system, the frequency res-
ponse can be measured only up to a certain distance. In this case a measurement of
up to 200 MHz was easily possible. The 3 dB bandwidth is found simply by deter-
mining the point at which the electrically measured transmission function has
dropped by 6 dB, here approximately 150 MHz.
Apart from the values actually measured, an approximation with a Gaussian
low pass function has been entered into the figure. By determining the frequency
f0 it is then possible to determine the bandwidth even when the measurement is not
possible because of the limited dynamics or bandwidth of the measuring system.
Figure 2.88 shows the measured transmission functions for a SI-POF and a
DSI-POF of 50 m length each. The optical 3 dB bandwidth for an SI-POF is
approximately 67 MHz, corresponding to a bandwidth-length product of
33 MHz · 100 m, with the NA of the fiber being 0.52. It follows that the measured
value is substantially greater than had been theoretically expected (approximately
14 MHz · 100 m, see Fig. 2.31). For DSI-POF (AN = 0.30) the measured value is
130 MHz, corresponding to 65 MHz · 100 m, with the theoretical value being
42 MHz · 100 m.
The measurement was carried out with a 520 nm LED. The LED had a wide
emission angle so that approximate equilibrium mode distribution can be assumed.
0
rel. power [dB] 50 m
DSI-POF
-3
-6
50 m
St.-NA-POF
-9
frequency [MHz]
-12
1 10 100 1,000
Fig. 2.88: Bandwidth measurement for SI-POF and DSI-POF
When measuring bandwidth, a two to four ranging factor of deviation from the
theoretical value of an ideal SI fiber can be generally expected, even when wor-
king in an EMD condition. The reason for this is the combination of mode depen-
dent attenuation and mode coupling described in Chapter 1. As a result of the con-
tinuous energy exchange that takes place between the faster and slower modes, the
delay does not rise in proportion to the length. The increased attenuation of those
beams having a particularly large propagation angle - many reflections at the
cladding - has the additional effect of reducing the pulse width.
Figure 2.89 shows the bandwidth measurement of a standard NA-POF for 3
different wavelengths for samples between 20 m and 100 m in length.
1000
bandwidth [MHz] bandwidth [MHz]
ACU 1000, Low-NA GH 4000, St.-NA

300
100
525 nm 525 nm
590 nm 590 nm
30
650 nm 650 nm
fiber length [m] fiber length [m]
10
10 20 50 100 10 20 50 100
Fig. 2.89: Bandwidth of a SI-POF and a DSI-POF at different wavelengths

Once again, the measurements in Fig. 2.89 were carried out with an LED
having an emission characteristic near to EMD (see [Gor98] and [Rit98]). The
figure reveals 2 significant items of information:
¾ The bandwidth of the POF does not decrease in proportion to the length-1; its
decrease is less than proportional.
¾ The bandwidth of the POF is nearly identical for the 3 attenuation windows.
2.5.3 Experimental Bandwidth Measurements
The section on experimental bandwidth measurements first summarizes the results

from earlier technical literature. As we shall see, determining the bandwidth is
among the most difficult metrological challenges with thick-core fibers. A series
of systematic measurements on the most diverse fibers is introduced as a supple-
ment to the first edition (Alexander Bachmann is responsible for the bandwidth
measurements at the POF-AC, cf. for example [Bun02a] and [Ziem04a]). Since
there are still no standards for the definition and measurement of bandwidth, the
following description will therefore not represent the definitive assessment.
2.5.3.1 Bandwidth of SI-POF

After presenting some examples of our own measurements in the previous section
we will now compare these with measurements carried out by other authors. Some
of the first efforts made to systematically investigate the bandwidth of SI-POF
were undertaken by [Tak91], [Tak93] and [Rit93]. As shown in Fig. 2.90, the
bandwidth was measured for SI-POF having lengths between 20 m and 100 m.
The fiber used was an EH4001 by Mitsubishi with an NA of 0.47. The bandwidth
was measured through the pulse broadening of a 660 nm laser pulse (150 ps).
Different launching devices were used to vary the NA between 0.10 and 0.65. The
detector used consisted of a wide area photo multiplier.
bandwidth [MHz] launch

2,000
conditions
1,000 AN Launch = 0.10
500 AN Launch = 0.65
theory
200
100
50
20
fiber length [m]
10
10 20 50 100
Fig. 2.90: Bandwidth measurement according to [Tak91]
The results allow three significant conclusions:

¾ The bandwidth of a SI-POF is always significantly higher than the theoretical
values for a SI-POF under theoretical UMD conditions, even with full laun-
ching in the acceptance area.
¾ The measured bandwidth is strongly dependent on the launching conditions.
¾ Although the bandwidth difference for measurements using different laun-
ching conditions decreases with the increasing length of fiber, it is still clearly
in evidence even after 100 m.
Figure 2.91 shows measurements of bandwidths for different detector NA, also
taken from [Tak91]. In principle, detection with a small NA means the same limi-
tation in mode number as launching light at a small angle so that it is not surpri-
sing to find that the value curves are similar.
1,000
bandwidth [MHz]
receiver detection
angle range
300
AN Det = 0.22
AN Det > 0.65
100
theory
30
10 fiber length [m]

10 20 50 100
Fig. 2.91: Bandwidth measurement according to [Tak91] with different receiver NA
Another measurement of bandwidth on standard NA-POF (1 mm) is presented

in [Tak93] (Fig. 2.92). Again, the measurement was carried out using the pulse
method at 650 nm. Apart from measuring the bandwidth, the half far field width
following the corresponding sample length was also determined.
The bandwidth was calculated from the far field width as follows:
2
A N, FF C
't mod and therefore : B z (C const.)
2nc 't mod
whereby tmod is the modal pulse propagation and B · z is the product of bandwidth
and length. Parameter C is a free selectable constant which depends on the coup-
ling conditions. The speed of light is c. In the formula AN , FF is not the fiber para-
meter indicated, but the value measured depending on length.
For a sample length of 10 m, the difference between the measured bandwidth
for launching the light with AN = 0.10 and AN = 0.65 is more than one order of
magnitude. For lengths up to 100 m this factor decreases to 2.
10,000 launching
bandwidth [MHz]
conditions
3,000
AN Launch = 0.10
1,000 AN Launch = 0.65
theory based on
300 far field width
100
30 fiber length [m]

10 20 50 100
Fig. 2.92: Bandwidth according to [Tak93]
When launching light with a small NA, the bandwidth drops disproportionately,
from approximately 80 MHz km to approximately 16 MHz km. This suggests
an increasing filling out of modal field. By comparison, when launching with a
large NA, the bandwidth is reduced somewhat more slowly than the length, from
approx. 4 MHz km to approx. 5 MHz km. This is due to the effect of mode
coupling and mode related attenuation.
The bandwidth values determined by means of the far field width correlate very
well with the results of the bandwidth measurements made by pulse propagation.
This suggests that mode dependent attenuation and mode conversion are the deter-
mining processes because they affect the bandwidth by changing the mode distri-
bution. In contrast, if mode coupling were more pronounced, the bandwidth would
also change without affecting the far field. However, any estimated quantification
based on these measured results alone would be questionable. In [Rit93] measured
results for the bandwidth of standard NA-POF at launching conditions of
AN = 0.10 and AN = 0.65 (Fig. 2.93) are also shown.
5,000 launching
bandwidth [MHz]
2,000 conditions
AN Launch = 0.10
1,000
AN Launch = 0.65
500
200
100
50
fiber length [m]
20
10 30 100 300 1,000
Fig. 2.93: Measured bandwidth of a SI-POF according to [Rit93]
Here too, the measured bandwidth for short lengths (20 m) differs by more than
an order of magnitude. For large lengths the difference is reduced corresponding-
ly. The authors calculate the bandwidth based on their own theory that follows the
concept of the diffusion model. Instead of investigating separate modes, this
model investigates modal groups that differ in their 2 angles of propagation (radial
and azimuthally).
The coupling between the modes is described by a diffusion constant that only
takes into account the energy transfer in neighboring mode groups. The model
also takes into account mode dependent attenuation.
In this work the remaining deviation between theory and measured values is
explained by means of the mechanism of mode coupling. In variance to the model,
this is a factor that is not independent of the angle. Simulations provide good
results if elongated scattering centers of 37 μm length and 2.5 μm diameter are
assumed in the fiber with random distribution and orientation along the axis of the
fiber (caused by the drawing process), as shown schematically in Fig. 2.94.
scattering centers
Fig. 2.94: Model for scattering centers in POF
An indication of a non-uniform inner structure of the PMMA fiber is the photo

(from >Fei00@) of the surface of a cut POF taken by a scanning electron micros-
cope and shown in Fig. 2.95. The fibril-like structures in the sub-Pm range can
clearly be seen.
Fig. 2.95: Microscopic structure of a PMMA POF cut ([Fei00])

Figure 2.96 shows further experimental results for the bandwidth of polymer
optical fibers [Kar92]. In each case collimated light or light with an angle adapted
to the fiber's NA (UMD) was launched into the POF. As was the case in the results
previously shown, very large differences result for short lengths of fibers. The
parameter shown in the figure here is the product of bandwidth and length.
bandwidth length product [MHzkm] launching

50
conditions
collimated
20 1.0 mm
0.5 mm
10
UMD
0.5 mm
5
1.0 mm
sample length [m]
2
5 10 20 50 100 200 500
Fig. 2.96: Measured POF bandwidth of a SI-POF according to [Kar92]
Apart from the effect of the launching NA, [Kar92] also investigates whether
the bandwidth depends on the size of the launched beam. In fact, for UMD laun-
ching, a larger bandwidth was found as well as a smaller light spot, compared with
complete illumination of the fiber cross-section; however, the differences are not
as pronounced as when the launch angle is changed. For collimated light the
relationship is reversed.
Because all processes described up to this point are only dependent on the
angle, it seems surprising to find that the size of the launching spot has an effect
on the measured bandwidth of SI fibers. However, when considering the fact that
mode conversion can cause deviations in location and deviations in angle after just
a short length of the specimen (see schematic in Fig. 2.97), the result becomes
understandable [Kar92].
bending
deviation in location
deviation in angle
Fig. 2.97: Conversion of spatial and angular distances

In [Poi00] the results of [Kar92] are compared with current measurements on 2

standard NA-POF by Toray and Mitsubishi (Fig. 2.98). These measurements
qualitatively confirm the previous results. For very short lengths of samples the
differences between small and large launching angles are even greater.
30,000 bandwidth [MHz]

NALaunch:
Toray 0.09
10,000 Mitsubishi 0.09
[Kar92] collimated
Toray 0.64
Mitsubishi 0.64
3,000
[Kar92] UMD
1,000
theory
300
100
length [m]
30
1 3 10 30 100 300
Fig. 2.98: Measured bandwidth of different SI-POF according to [Poi00]
2.5.3.2 Bandwidth Measurements on SI-POF

This section as well as the following four sections deals with bandwidth measure-
ments conducted at the POF-AC Nürnberg. All measurements were carried out
under uniform measurement conditions.
Semiconductor diodes with a wavelength of 650 nm or 850 nm respectively
served as transmitters. Both lasers can be modulated analogously up to 2 GHz. A
singlemode glass fiber is mounted firmly to the laser diodes. Using a combination
of different microscope lenses and optical apertures the coupling angle in the area
AN Launch = 0.01 to 0.64 can be varied. The coupling spot is directly visible through
a beam splitter. With the aid of adjustment screws the size as well as the position
of the light spot can be changed. Figure 2.99 shows the complete setup of the
measurement device.
A commercial product on the basis of a 400 μm Si-PIN photodiode with an
integrated preamplifier and about 1.5 GHz bandwidth was used as a receiver. In
order to attain mode independence the receiver was connected to a 1 mm mixed
glass fiber bundle with a large NA.
Fig. 2.99: Experimental setup of bandwidth measuring by POF-AC
The lengths and NA-dependent bandwidths were systematically measured for a

series of different step index profile fibers. An overview can be found in
[Bun02a]. The following Figs. 2.100 to 2.102 show the results for three types of
fiber:
¾ 1 mm standard PMMA POF with AN = 0.46
¾ 1 mm standard POF made of cross-linked PMMA with AN = 0.54
¾ 1 mm polycarbonate POF with AN = 0.75
B3dB [MHz]
5,000 NAlaunch:
2,000 0.64
0.48
1,000
0.33
500 0.19
0.09
200
0.05
100
50
20 length [m]
5 10 20 50 100
Fig. 2.100: Bandwidth measurement of a 1 mm SI-PMMA-POF
For a 1 mm PMMA POF (Toray PFU CD1000, see also [Ziem04a]) 3 dB band-
widths for lengths between 5 m and 100 m were measured. The coupling angle
was changed for NA values between 0.05 and 0.65 with the unit described above.
For short fiber lengths the bandwidths measured differ by almost a magnitude
which demonstrates once again the importance of correct measurement conditions
for correctly indicating the bandwidth values. After a 100 m test length there still
is a factor of two between the values measured. The curves for under filled launch
(small NA) fall more steeply than with length caused by a predominance of mode
mixing. For overfilled launch (large NA) the curves run flatter. Here the mode-
dependent attenuation dominates. The next figure shows the results with a 1 mm
POF made of modified PMMA (Toray PHKS CD1001). The fiber is specified
with a NA of 0.54.
B3 dB [MHz] NAlaunch:
3,000
AN = 0.05
AN = 0.09
1,000
AN = 0.19
300 AN = 0.33
AN = 0.48
100 AN = 0.64
30
5 10 20 50 100 length [m]
Fig. 2.101: Bandwidth measurement of a 1 mm SI-mod. PMMA-POF
Since the losses of this fiber lie at about 300 dB/km at 650 nm, test lengths of
only up to 50 m could be measured. Incidentally, the measurement results are
similar to a large degree to the results of the PMMA POF.
3.000 NAlaunch:
B3 dB [MHz]
AN = 0,05
1.000 AN = 0,09
AN = 0,19
300 AN = 0,33
AN = 0,48
100 AN = 0,64
length [m]
30
1 2 5 10 20
Fig. 2.102: Bandwidth measurement of a 1 mm SI-PC-POF
The third fiber tested is the polycarbonate POF FH4001 from Mitsubishi. The
NA of the fiber lies at 0.75, the attenuation amounts to 650 nm at about
800 dB/km, whereby the maximum measurement length remains limited to 20 m.
Surprisingly, the bandwidth differences between the three types of fiber are
only very slight although there were clear differences in the NA. One explanation
for this could be the greater effects for mode mixing and above all for the mode-
dependent attenuation which occurred in the fibers made of modified PMMA and
polycarbonate. Figures 2.103 and 2.104 illustrate the far fields of the three fibers
in comparison (cf. [Bun02a]).
B3 dB [MHz]
3.000
PC
1.000 PHKS
PMMA
300
NALaunch = 0.33
100
length [m]
30
2 5 10 20 50 100
Fig. 2.103: Comparison of the bandwidths of different SI-POF
1000
power [a.U.]
mod. PMMA
800
PC
600
PMMA
400
200
T [°]
0
-40 -30 -20 -10 0 10 20 30 40
Fig. 2.104: Comparison of the farfields of different SI-POF
The fibers of PMMA and PC - each after 10 m - have half-value widths of

about 27°. The fibers of modified PMMA have only 17°. Here the share of mode-
dependent attenuation predominates over the nominally larger NA.
Fig. 2.105: Comparison of the far fields of different SI-POF (3-d representations)
As part of the European Project POF-ALL (www.ist-pof-all.org) other compre-

hensive measurements of both length and launch-dependent bandwidths of diffe-
rent fibers were carried out. The following Figures 2.106 and 2.107 show the
measurement results for a 1 mm standard POF (Luceat, high quality fiber) and for
a 500 μm standard POF.
10,000
B3 dB [MHz] NALaunch
5,000 0.05
0.10
2,000
0.19
1,000 0.34
0.47
500
0.65
200
100
50
1 2 5 10 20 50 100
fiber length [m]
Fig. 2.106: Bandwidth measurements of a 1 mm SI-POF (Luceat, HQ)
Both fibers essentially show comparable results. Since the fibers also have very
similar attenuation values they can be used in almost all the same applications.
The advantages of the thinner fibers are primarily the smaller space needed, an im-
portant point with multiple cables, and the smaller bending radius. The argument
that the fibers with a smaller core diameter would enable higher bit rates or better
receiver sensitivity because of the smaller photodiodes has for the most part since
been dropped because of technical developments.
10,000
B3 dB [MHz] NALaunch
5,000
0.05
2,000 0.10
0.19
1,000 0.34
500 0.47
0.65
200
100
50
1 2 5 10 20 50 100
fiber length [m]
Fig. 2.107: Bandwidth measurements of a 0.5 mm SI-POF
2.5.3.3 Bandwidth Measurements on MC- and MSI-POF

Multicore and multistep index profile POFs allow significantly greater bandwidths
than conventional step index profile fibers. In the case of MC-POFs the diffe-
rences in propagation time between the different individual cores, in addition to
mode dispersion, have to be added. The pure length differences, however, should
hardly play a role. The pure path differences alone between the modes amount to
about 6% at a maximum propagation angle of 20° in the fiber. Since the fibers lie
well-ordered in the MC-POF the geometric differences in length lie at the most in
the thousandth range.
Of greater significance is the fact that the fibers in the MC-POF are deformed
in different ways. For example, differences among the fibers in the middle and at
the edge of the bundle can already be seen in the attenuation. These differences
are also formed for the mode selective processes resulting in different average
propagation speeds in the individual cores.
launching with magnified light spot: launching with mode field converter:
- medium fibers obtain small angles only - all fibers obtain around the same
- outer fibers obtain large angles optical power and rays of all angles
Fig. 2.108: Optimal launching into multicore fibers

In order to register these effects when making a measurement, a so-called mode

field converter (MFC) is used. The coupling unit shines into a short piece of SI
fiber with a large NA. The far field distribution remains intact as the light is distri-
buted over the fiber cross-section. This ensures that the individual fibers receive
approximately identical light intensity and comparable angle distributions. The
difference between this arrangement as opposed to a simple widening of the light
spot is depicted schematically in Fig. 2.108.
As indicated above, there are in the meantime different MC-POFs. Here we
wish to present the results of the bandwidth measurements of two 1 mm MC-POFs
with 37 cores (Fig. 2.109) and 217 cores (Fig. 2.110, see also [Ziem02a]).
2,000
B3 dB [MHz] launch NA:
PMC 1000
AN = 0.09
37 cores
AN = 0.19
1,000 AN = 0.33
AN = 0.48
AN = 0.64
500
200
length [m]
100
20 30 40 60 80 100
Fig. 2.109: Bandwidth measurements of a 37-cores MC-POF (measured on single samples)
2,000
B3 dB [MHz] MCS 1000 launch NA:
217 cores AN = 0.09
AN = 0.19
1,000
AN = 0.33
AN = 0.48
500 AN = 0.64
200
100
20 30 40 60 80 100
length [m]
Fig. 2.110: Bandwidth measurements of a 217-core MC-POF (measured on single samples)

Both fiber types show considerably greater bandwidth values compared to

standard SI-POF. The 37 core fiber above all shows hardly any drop in the band-
width over great lengths, especially since the dependence on the coupling condi-
tions is very small. The reason is the very great mode dependence of the attenu-
ation. This fiber possesses a double step index structure. Using a laser coupling it
has been possible to transmit 1 Gbit/s over this fiber for 100 m.
Systematic investigations of the bandwidth have been carried out at the
POF-AC on a 37 core POF sample with a relatively small diameter (400 μm). The
results for the two different NA are shown in Fig. 2.111. The bandwidth of this
fiber is almost independent of the launch conditions. The reason for this is the
strong mode-dependent attenuation which, as described above, occurs intensively
with very thin fibers and leads to a equilibrium mode distribution after very short
lengths.
B3 dB [MHz]
10,000
NALaunch
0.05
5,000
0.65
2,000
1,000
500
200
0.1 0.3 1 3 10 30 100
fiber length [m]
Fig. 2.111: Bandwidth measurements of a MC-POF (measurements on a single fiber
sample, cut-back method)
Multi step index fibers have already been introduced by different manufac-
turers. However, they are not yet ready to go into mass production. The youngest
product so far is the ESKA MIU from Mitsubishi-Rayon, a fiber with three diffe-
rent layers. Using a sample length of 100 m of this fiber, a bandwidth of almost
300 MHz was ascertained. Figure 2.112 shows the frequency response.
3
rel. level [dB]
0 NA 0.10
NA 0.34
-3 NA 0.64
-6
-9
-12
-15
-18
-21
-24
10 20 50 100 200 500 1000
frequency [MHz]
Fig. 2.112: Frequency responses of a 100 m MSI-POF
2.5.3.4 Bandwidth Measurements on GI-POF

The bandwidth measurements of graded index profile fibers are associated with a
series of particular difficulties. First of all, the attenuation is relatively high with
polymer fibers so that the sample lengths can not be very great because of the
limited dynamics of the measuring system. For PMMA POF the maximum
measuring lengths lie between 50 m to 100 m. For PF GI-POF lengths of some
100 m can be used. On the other hand, POFs have a relatively large core diameter.
The detector must be relatively large in order to record most all modes and thus
obtain a meaningful bandwidth measurement. Then again the size of the diode
limits the bandwidth of the detector. The only commercially available measuring
system that can be used for this special task is the optical oscilloscope from
Hamamatsu (described in the Chapter on measurement techniques). However,
only measurements in the time domain are possible.
The bandwidths of PMMA GI-POF and PF-GI-POF have been measured at the
POF-AC. The transmission functions for the PMMA GI-PF OM-Giga from Opti-
media (see also [Yoo04], [Rich04]) and a PF-GI-POF (Nexans, see [Gou04]) are
illustrated in Fig. 2.113 and 2.114.
The optical bandwidth of the fiber at 1.504 MHz was ascertained by adapting
the measuring curve to a Gaussian function. This value, however, can clearly
fluctuate with slightly altered launch conditions. The measuring conditions lie
close to the “worst case scenario”. When there is under-launching, e.g. with a
VCSEL transmitter, even greater values can be attained.
rel. electr. level [dB]

0
-1
-2
-3
50 m OM-Giga
-4 OLD = 650 nm
ANLaunch = 0.34
f3 dB opt. = 1,504 MHz
-5
frequency [MHz]
-6
10 20 50 100 200 500 1,000
Fig. 2.113: Frequency response of 50 m OM-Giga (AN = 0.34; 650 nm)
In order also to be able to measure bandwidths of several GHz with thick core
fibers, an optical oscilloscope is a practicable device, whereby the widening of a
short laser pulse (about 120 ps) is measured. In [Lwin06] the results for the
OM-Giga are shown compared with the microstructured POF (with effective
graded index profile).
350
pulse broadening [ps]
300
Optimedia 1,000 μm
250
200
MPOF: 500 μm
150
length [m]
100
15 25 35 45 55 65
Abb 2.114: Pulse broadening measurement of a MPOF and GI-POF
A pulse width of about 340 ps corresponds approximately to an optical band-

width of 1.4 GHz. The value matches pretty much the measurements in the fre-
quency range when taking the various problems of measurement techniques with
such large frequencies into account.
The next illustration shows the frequency response for a PF-GI-POF at the
wavelengths 650 nm and 850 nm together with the fitted Gaussian functions.

+1
measurement 650 nm
Gaussian fit 650 nm
0 measurement 850 nm
Gaussian fit 850 nm
-1
-2
-3
300 m PF-GI-POF
-4 OLD = 650 nm/850 nm
ANLaunch = 0.10
-5
-6
0 200 400 600 800 1000 1200
frequency [MHz]
Fig. 2.115: Frequency response of a PF-GI-POF
The 3 dB bandwidths are around 1,600 MHz for both fibers. The bandwidth-
length product is at about 500 MHz · km, somewhat in the range of conventional
multimode graded index glass fibers (cf. further results in [Bach01]).
2.5.3.5 Bandwidth Measurements on MC-GOF and PCS

The bandwidth measurement of multimode glass fibers proceeds according to the
same principles. First, the glass fiber bundles from the Schott manufacturing com-
pany are measured. The bundle consists of about 400 individual fibers each having
a diameter of 53 μm. The fibers have been hot pressed into the plug so that the
overall diameter is about 1 mm.
In Fig. 2.116 the frequency response for various coupling conditions are shown.
The NA of the fiber lies at 0.50. The fiber bandwidth does not recognizably
change when coupling in at great angles.
With full mode launch the measured bandwidth amounts to about
150 MHz · 20 m which is almost exactly the same value as for SI-POF with a
comparable NA. Consequently, this fiber type can be used alternatively to the
1 mm St.-NA POF when either high temperatures or very tight bending radii are
necessary ([Lub04b]).
2
0
-2
20 m fiber
-4
at 650 nm
-6
launch NA:
-8
AN = 0.10
-10 AN = 0.34
AN = 0.46
-12
AN = 0.60
-14 AN = 0.64
f [MHz]
-16
10 20 50 100 200 500 1000
Fig. 2.116: Bandwidth measurement of 20 m MC-GOF
In another series of measurements the length dependence of the bandwidth of

MC-GOF was investigated. Fig. 2.117 shows the results for 3 different launch
conditions. The bandwidth decreases almost linearly with the length so that one
can assume that the influence of mode mixing is relatively small.
B3 dB [MHz]
3,000 1 mm MC-GOF
375 cores
NAfiber: 0.50
O = 650nm
1,000
launch NA:
AN = 0.64
300
AN = 0.34
AN = 0.10
length [m]
100
2 5 10 20 50
Fig. 2.117: Length- and NA-dependent bandwidth of MC-GOF
Finally, the bandwidth for different lengths was determined using a 650 nm
laser. In order to be able to make measurements relatively independently of mode,
a 1 m long SI-POF was used as an adapter fiber at both the transmitter and the
receiver. Figure 2.118 shows the results.
1000
B3 dB, opt. [MHz] excitation by laser
NAlaunch | 0.30
O = 650 nm
500
200
length [m]
100
10 20 30 40 50 60
Fig. 2.118: Bandwidth of a MC-GOF excited by a laser source
This type of fiber is suitable for the transmission of data rates in the Gbit/s
range over lengths of 10 m to 20 m.
Another glass fiber version which has gained increasing attention is the PCS,
i.e. silica glass fibers with a polymer cladding. The typical NA lies around 0.37.
However, there are versions available with a NA up to 0.48. Accordingly, the
bandwidth of PCS should lie in the range of DSI-POF. At the POF-AC predo-
minantly fibers with a core diameter of 200 μm - the most commonly used value -
were measured. In Fig. 2.119 the length and launch-dependent results for a typical
PCS are represented. The fiber, 200/230 μm with a 500 μm primary coating, was
laid out for this measurement as a loose bundle with a diameter of about 30 cm
(see also [Ziem04a]).
B3 dB [MHz]
2000
launch NA:
1000 0.02 0.26
0.09 0.34
500
0.17 0.46
200
100
50 200 μm PCS
loose bundle
length [m]
20
10 20 50 100 200 500
Fig. 2.119: Bandwidth of a 200 μm PCS
This fiber was specified with a bandwidth of 100 MHz · 100 m. This value can
be achieved for an under filled launch. For a full launch, however, you can only
attain about 60 MHz · 100 m. The differences between the different launch condi-
tions hardly decreases with fiber lengths up to 250 m. Mode mixing hardly occurs
with this measurement. The measurement was repeated for the same type of fiber,
whereby the fiber was wound around a spool. The results are shown in Fig. 2.120.
2000
bandwidth [MHz] launch NA:
1000 0.02 0.26
500 0.09 0.34
0.17 0.46
200
100
50 200 μm PCS
fiber on a spool
length [m]
20
10 20 50 100 200 500
Fig. 2.120: Bandwidth of a 200 μm PCS
The results pretty much agree for short fiber lengths. For longer lengths, how-
ever, the differences roughly disappear between the different launch conditions for
the rolled up PCS. This can only be explained by a recognizable increase in the
mode mixing. The bandwidths dependent on the coupling NA are compared for
250 m long samples in Fig. 2.121.
90
Bopt, 3 dB [MHz] 250 m PCS
80
70
60
50
40 fiber on a spool
30
20 loose bundle
10
launch NA
0
0.00 0.10 0.20 0.30 0.40 0.50
Fig. 2.121: Bandwidth comparison of 250 m PCS
Examinations of the different types of 200 μm SI-PCS confirm the measure-

ments mentioned above. The specified bandwidth-length products of 10 to
20 MHz · km could be attained for all fibers examined only with under filled
launch. Unfortunately, none of the currently active manufacturers provided any
data on measurement conditions for the bandwidths indicated. Even the corres-
ponding standards are completely missing. Should PCS seriously advance into
areas of application in which the available bandwidth is to be completely used,
then a lot of work still has to be done in this area. A comparison between the
launch-dependent bandwidths is represented in Fig. 2.122.
BL [MHz km]

40
30
20
15
10
8
4
0.0 0.1 0.2 0.3 0.4 0.5
launch NA
Fig. 2.122: Bandwidth dependence on launch conditions for 5 different PCS types
A number of publications, especially in regard to applications in the Gigabit-

Ethernet and 10Gigabit-Ethernet ranges, exist for a bandwidth of 50/125 μm GI
glass fibers. Articles providing an overview include [Oeh02] and [Bun03a].
With GI glass fibers, too, keeping the exact parabolic index profile as well as
the mode-selective coupling play the most important role in achieving large band-
widths.
Originally, two different types of fibers were specified:
¾ customary in the USA: 62.5/125 μm fiber (AN = 0.275 ± 0.015)
¾ customary worldwide: 50/125 μm fiber (AN = 0.200 ± 0.015)
The typical bandwidth-length product is 160 MHz · km to 200 MHz · km
(62.5 μm) when using an 850 nm LED as emitter. 500 MHz · km is attained with
1,300 nm laser emitters. The limiting factor is the refractive index dip in the
middle of the fiber which is caused by the production technology.
For fast Ethernet (125 Mbit/s) the bandwidths entirely suffice to bridge distan-
ces of up to 1 km. The first range limitations (maximum of 275 m at 850 nm emit-
ters and 62.5 μm fiber) arise with Gigabit-Ethernet so that a new class of fibers
(OM2) has been defined which generally guarantees a transmission range of
550 m.
In the worst case a data rate of 10 Gbit/s could be transmitted on OM1 fibers
over about only 30 m. OM2 fibers are also limited to about 80 m. In order to be
able to transmit high data rates, three different procedures have been suggested:
¾ Splitting the data rate into 4 × 2.5 Gbit/s which are then transmitted by
WDM on a fiber.
¾ Emitter with so-called Restricted Mode Launch (RML) or Effective Laser
Launch (EL) respectively, whereby the power is coupled if possible within
the annulus with a diameter of between 4.5 μm and 19 μm. Moreover, the
NA of the emitter may not be too large.
¾ Use of the new OM3 fiber class which has been optimized for the employ-
ment of 850 nm VCSEL.
An overview of the specified characteristics of the different GI-GOFs is
presented in Table 2.14. Specific products can on occasion clearly surpass these
parameters.
Table 2.14: Properties of MM-GI glass fibers
OM3
Class Unit OM1 OM2 OM3
550m
Fast Gigabit 10Gbit 10Gbit
typical applications
Ethernet Ethernet Ethernet Ethernet
core- [μm] 50/62.5 50/62.5 50 50
D at 850 nm [dB/km] 3.5 3.5 3.5 3.0
D at 1.300 nm [dB/km] 1.5 1.5 1.5 1.0
BW 850 nm (OFL) [MHzkm] 200 500 1,500 3,500
BW 1.300 nm (OFL) [MHzkm] 500 500 500 500
BW 850 nm (LD) [MHzkm] n.d. n.d. 2,000 4,700
(OFL: Overfilled Launch)
Other less customary fiber types are, for example, GI-GOF with a core dia-
meter of 100 μm and a cladding diameter of 140 μm. Fig. 2.123 shows the fre-
quency response of a 500 m long sample with three different launch conditions. At
200 MHz · km the results lie in the range of the fiber specifications.
The last fiber presented here is the semi-GI-PCS described above. The
measurement conditions become extremely more noticeable here so that the
measurement results shown may not be conclusively representative.
1
0
-1
500 m fiber
-2 at 650 nm
-3
-4
-5 launch NA:
AN = 0.10
-6
AN = 0.34
-7 AN = 0.64
-8
1 10 100 f [MHz] 1000
Fig. 2.123: Frequency response of a 100 μm GI-GOF
Fig. 2.124 first shows the frequency response with a 500 m long sample for 6
different launch conditions measured at a wavelength of 650 nm.
rel. opt. power [dB]

0
AN Launch =
-2 0.03 .. 0.64
-4
-6 500 m
Semi-GI-PCS
-8
-10
-12
1 3 10 30 100 300
frequency [MHz]
Fig. 2.124: Bandwidth of a Semi-GI-PCS
The bandwidth-length product of the fiber was determined as having values

between 24 and 55 MHz · km which clearly lies above the specification of
100 MHz · km. Bandwidths with their length and launch dependence are also
determined for this type of fiber. Fig. 2.125 summarizes the results.
3000 B3 dB [MHz]
launch NA:
AN = 0.02
1000 AN = 0.09
AN = 0.17
AN = 0.26
300
AN = 0.34
AN = 0.46
100
length [m]
30
10 20 50 100 200 500 1000
Fig. 2.125: Bandwidth measurement of Semi-GI-PCS
What is striking is the low dependence of the bandwidth on the launch con-
ditions with longer sample lengths. Evidently, there is a significant exchange of
energy between the SI and GI modes in the fiber. The specified bandwidth value
could only be determined in short fiber lengths with under filled launch.
Bandwidth measurements on semi-GI PCS have also been published by
[Aiba04] and [Aiba05], whereby a method was used in which a light pulse circu-
lates in a 100 m long ring and passes an acousto-optic modulator after every pass.
The numerical aperture of the coupling optics amounts to only 0.25 and SI modes
are for the most part suppressed. The results for the frequency response, deter-
mined by Fourier transformation, are shown in Fig. 2.126.
rel. opt. power [dB]

0
-2 1st circulation
-4 10th circulation
-6
-8
f [GHz]
-10
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Fig. 2.126: Frequency responses of a Semi-GI-PCS according to [Aiba04]
The bandwidths thus determined are shown in Fig. 2.127. The values lie higher
by a factor of ten than the values measured with full launch on long fibers. This
reveals impressively how important correct specifications of the measurement

conditions are when indicating bandwidth values.
The succinct statement in the Sumitomo data sheet that “the bandwidth can
change under other measurement conditions” is of little help.
B3 dB, opt. [GHz]

2.0
semi-GI -PCS
1.5
1.0
0.8
0.6
0.4
length [m]
0.3
100 200 400 600 1000
Fig. 2.127: Bandwidth of a Semi-GI-PCS according to [Aiba04]
2.5.3.6 Comparison of Bandwidth Measurements and Calculations

The diverse measurements of fiber bandwidths show that the same principles are
essentially valid for thick glass and polymer fibers. Important effects are:
¾The bandwidth drops with the square of the numerical aperture by increasing
the differences in propagation time among the individual modes.
¾The diameter of the fiber does not play any role in regard to the bandwidth.
¾Strong mode-dependent attenuation increases the bandwidth of fibers, but it
also leads to a rise in transmission losses.
¾Multicore fibers and fiber bundles permit smaller NAs with the same bending
radius and thus greater bandwidths.
¾The bandwidth of fibers greatly depends on the launch and detection condi-
tions. The difference can be » 10 for short fiber lengths. When stating the
bandwidth in data sheets, measurements should always be made with UMD
(full launch) or EMD (equilibrium mode distribution).
¾Graded index profiles increase the bandwidth up to two magnitudes. How-
ever, the index profile must be as ideal as possible - it should be parabolic
when the chromatic dispersion is disregarded.
¾In the case of a non-ideal GI profile a large bandwidth can still be attained
through a selective launch.
¾In addition, the chromatic dispersion especially with glass GI fibers has to be
taken into account (this will be discussed in the next paragraph).
¾It is technically easier to produce a multi-stepped index profile, with which
the bandwidth can clearly be increased, than a GI profile.
¾Semi-GI fibers have large bandwidths, above all over short lengths and when
coupling into small angles.
¾The bandwidth of individual fibers - not yet placed in cables - under labora-
tory conditions can depend to a great extent on the external conditions, depen-
ding on the degree of induced mode coupling.
A comparison between POF and PCS is particularly interesting since both can
be used alternatively in many applications. The length-dependent bandwidths of
both types of fiber with full launch are illustrated in Fig. 2.128.
B3 dB, opt. [MHz]

1.000
PCS, NALaunch = 0.48
POF, NALaunch = 0.64
300
100
length [m]
30
3 10 30 100
Fig. 2.128: Bandwidth comparison of POF (fiber-NA: 0.50) and PCS (NA: 0.37)
Theoretically, the PCS should show about 50% greater bandwidth because of
its smaller NA - which has just about been confirmed by measurements. Both
measurement curves run approximately parallel which suggests similar magni-
tudes in mode-dependent processes. The angle-dependent attenuation of a typical
PCS fiber is illustrated in Fig. 2.129.
225 excess loss [dB/km]

200 fiber length
50 m
175 100 m
150 200 m
125
100
75
50
25
0 T[°]
-25 -20 -15 -10 -5 0 5 10 15 20 25
Fig. 2.129: Mode dependent loss of a PCS (at 650 nm)
400
attenuation [dB/km]
300 50 m
200
100 m
100
11 dB/km
T [°]
0
-30 -20 -10 0 10 20 30
Fig. 2.130: Mode dependent loss of a Semi-GI-PCS (at 650 nm)
PCS does indeed show very large mode-dependent attenuation, the intensity of
which is comparable to POF. This explains the similar behavior even if the core
material itself has a very much lower attenuation.
A schematic comparison of typical bandwidth values for the different multi-
mode fibers described above are illustrated in Fig. 2.131. The values, as already
mentioned several times, can clearly deviate for specific products or under diffe-
rent measurement conditions.
PC-POF Ø: 1000 μm
MC-GOF Ø: 1000 μm
St.-NA-POF Ø: 1000 μm
200 μm PCS Ø: 200 μm
DSI-POF Ø: 1000 μm
SI-MC-POF Ø: 1000 μm
Semi-GI-PCS Ø: 200 μm
DSI-MC-POF Ø: 1000 μm
MSI-POF Ø: 750 μm
OM-Giga Ø: 900 μm
GI-GOF OM1 Ø: 62.5 μm
GI-GOF OM2 Ø: 50 μm
PF-GI-POF Ø: 120 μm
GI-GOF OM3 Ø: 50 μm
OM3 mit LD Ø: 50 μm
1 10 100 1,000 10,000
bandwidth [MHz·km]
Fig. 2.131: Bandwidth comparison of different optical fibers (typical values)
The bandwidths of the fibers presented vary over more than 3 magnitudes. If
singlemode fiber is used, however, then nowadays there is practically no longer
any bandwidth limit. Mode dispersion no longer arises. Chromatic and polariza-
tion mode dispersion can be compensated for as one likes. The significance of
chromatic dispersion will be discussed in the next section.
2.5.4 Chromatic Dispersion in Polymer Optical Fibers
In all optical media we can observe the effect that the speed of propagation of
light of different wavelengths differs. When we differentiate the propagation con-
stants according to wavelength, we obtain the so-called chromatic dispersion,
usually expressed in ps/nm·km. This constant indicates by how much a signal's
delay will vary with the wavelength. In the typical application range of optical
fibers this value is negative which means that with increasing wavelength the
delay becomes smaller (corresponding to greater speed). Figure 2.132 shows the
chromatic dispersion for silica glass, PMMA and a typical fluorinated polymer
(according to [Koi97a]).
200
dispersion [ps/(nmkm)]
0
-200
-400
-600 PF-Polymer
-800 silica glass
PMMA
-1,000
-1,200
400 500 600 700 800 900 1,000 1,200 1,400 1,600
wavelength [nm]
Fig. 2.132: Dispersion of different materials
Typical semiconductor sources feature certain spectral widths that range from
some 10 nm for LED up to a few MHz for lasers (corresponding to some 10-5 nm).
In addition, there is the fact that when a light source is modulated there is always a
spectral broadening that cannot be less than a certain theoretical limit. This effect
only plays a role, however, with spectral singlemode lasers and with very high
data rates.
Figure 2.133 shows a schematic illustration of the effect of chromatic disper-
sion on a light pulse that has a given spectral width. A pulse with a certain spec-
trum of the width 'O is launched into the fiber. After passing through the fiber
(length L) and experiencing a certain amount of dispersion D, the pulse has the
width 'W = D · L · 'O, whereby the shorter wave components arrive first. (cf.
Fig. 2.38 as well).
spectral shape
of the source
'O
O
't = DL'O
POF
length L
t t
input pulse output pulse
broadening by time
Fig. 2.133: Influence of chromatic dispersion
For silica singlemode fibers, the value for chromatic dispersion at 17 ps/nm·km
lies within the range of the smallest fiber attenuation at 1,550 nm wavelength.
Today, DFB-laser diodes are predominantly used for long-distance systems, the
spectral width of which is a maximum of a few MHz. What matters here essen-
tially is the broadening effect that is brought about by the data itself. In this case,
1 nm corresponds to approximately 125 GHz of spectral width. This means that
for a data rate of 10 Gbit/s a spectrum in the range of one-tenth nm is generated.
Where the permissible bit broadening is 0.05 ns, the fiber link may have a length
of approximately 30 km. For 2.5 Gbit/s this value rapidly increases to approxima-
tely 500 km due to the narrower spectrum and the greater pulse broadening per-
mitted. Conventional 2.5 Gbit/s systems can operate without specific actions
against dispersion. However, all systems that have many inline fiber amplifiers or
higher bit rates require devices to counteract chromatic dispersion. The most
common method today is the use of dispersion compensating fibers with strong
negative dispersion. Since these fibers utilize waveguide dispersion they can only
be produced as singlemode fibers.
The situation is significantly different for POF. The chromatic dispersion of
PMMA-POF with over 300 ps/nm·km at 650 nm wavelength is over 20 times lar-
ger than of silica fibers at 1,550 nm wavelength. For POF it is also usual to use
LED with a typical spectral width of 20 nm to 40 nm and not lasers that have just
a few tenths of a nanometer of spectral width. On the other hand, there are the
typically short distances of POF systems and the moderate bit rates. Table 2.15
lists some examples for the effect of chromatic dispersion in POF systems.
Table 2.15: Influence of chromatic dispersion in POF systems

Example Bit Rate/ Wavelength/ Pulse Broadening/
POF-Length Spectr. Width rel. to the Bit Length
SI-POF 50 Mbit/s / 650 nm LED 0.375 ns
50 m 20 nm FWHM 2 % of the bit time
ATMF 155 Mbit/s / 650 nm LED 0.75 ns
DSI-POF 50 m 40 nm FWHM 12 % of the bit time
ATMF 155 Mbit/s / 525 nm LED 2.8 ns
DSI-POF 100 m 40 nm FWHM 43 % of the bit time
IEEE1394 500 Mbit/s / 525 nm LED 1.96 ns
MC-POF 70 m 40 nm FWHM 98 % of the bit time
STM16 2,500 Mbit/s / 650 nm LD 0.05 ns
PF-GI-POF 200 m 2 nm FWHM 12 % of the bit time
The first three examples are based on LED for transmitting data rates up to
155 Mbit/s over a maximum length of 100 m. Even in the unfavorable case of
using green LED, pulse broadening is less than one ½ the bit length so that there is
only a small effect on the system. In the fourth example, the intention is to trans-
mit an IEEE1394 S400 data stream (with 500 Mbit/s physical data rate) over a
distance of 70 m using a green LED. Here pulse broadening is nearly in the same
range as the bit length. When this deteriorating effect due to mode dispersion is
added, one can see that this system can only work with considerable additional
efforts. It may, for example, be possible to partially provide electrical compen-
sation, whereby higher optical receiving power is required. When using data rates
from ½ Gbit/s to 1 Gbit/s, the use of spectrally narrower sources becomes neces-
sary. These primarily include RC-LED and VCSEL (see Chapter 4), and for even
higher requirements DFB laser diodes. In most cases this selection is required
anyway due to the limited modulation bandwidth of LED.
Fluorinated graded index profile polymer fibers feature significantly reduced
chromatic dispersion compared with PMMA-POF. These fibers are designed for
use in Gbit/s systems operating at spectral ranges between 800 nm and 1,300 nm.
It is for these demands only that laser diodes can be considered, not least due to
the smaller core diameters, the spectral width of which is a few nanometers at
most. The last row shows that in such a case chromatic dispersion can be neglec-
ted even for a transmission length of a few 100 m.
2.5.5 Methods for Increasing Bandwidth
Generally, the theoretical bandwidth of polymer fibers is calculated on the basis of

two essential assumptions. One assumption is that the launch of light at the fiber
entrance takes place in uniform mode distribution and that the detector will re-
ceive all modes. The second assumption is that the attenuation of all modes is
nearly constant. However, in practice polymer fibers, and in particular step index
fibers, show completely different behavior. In the first place, it is relatively diffi-
cult to illuminate all modes uniform at the entrance of the fiber. In many cases
laser diodes are used where the emitting angle is significantly smaller than the
angle of acceptance of a SI-POF. The use of solid-state lasers or gas lasers, the
exact wavelength of which is often required for measuring purposes, is even more
problematical. These lasers emit collimated light so that only a small proportion of
the POF modes can be excited. When using glow lamps or discharge lamps, opti-
cal devices are used to collimate the light to the fiber. For this reason it is difficult
to find lenses that actually work with consistent efficiency in the given acceptance
range. All this has the effect in a concrete experiment of increasing the deviations
of the actual bandwidth in comparison with the theoretical limit value. This is a
very undesirable effect when attempting to define characteristics by making
measurements of this kind, as shown in Chapter 7. However, for high bit-rate data
transmission this situation can in practice also be beneficially exploited as shown
by the following examples.
Figure 2.134 demonstrates the most important methods for increasing the
bandwidth of a POF.
launch with detection with

small angle fiber without small angles
connectors, bends
IN and splices OUT
high pass filter for high pass filter for

dispersion dispersion
precompensation postcompensation
Fig. 2.134: Methods for increasing bandwidth (cf. p. 441 as well)
Launching light at a small angle as well as detecting just a selected angle range
has the effect of restricting the modes involved in signal transmission and thus
reducing pulse broadening. It is possible to electrically compensate for the resul-
ting low pass behavior, both before as well as after the POF link. To date the most
significant increases in bandwidth for a POF system have been described in
[Bat92] (see also [Bat96a] and [Yas93]). The following components were utilized:
¾ Launch with a small AN = 0.11, thereby exciting only a few modes with only
small differences in delay.
¾ Pre-distortion of the LD excitation signal (peaking); high pass (33 pF [[ȱ51 :).
¾ Detection with low NA (modes with large delay differences are blanked out).
¾ Dispersion compensation behind the receiver; high pass (8 pF [[ȱ200 :).
It was possible to transmit at more than 500 Mbit/s across a distance of 100 m
of standard NA-POF (see also chapter 6). However, all these measures are usually
at the expense of a reduced power budget, as summarized in Table 2.16.
Table 2.16: Consequences of different bandwidth increasing methods
Method Penalty for the Power Budget due to:

peaking lowering of source modulation depth
low NA launch decreasing POF coupled optical power for sources with
broad emission angle
low NA detection loss of light with high propagation angle at the fiber output
post compensation amplified noise at higher frequencies
It follows that the use of such methods is of particular interest in systems that
have adequate power reserves. POF attenuation across very short distances is
hardly of importance; on the other hand, the use of high data rates is of interest in
various applications. Chapter 6 will describe experiments for transmitting Gbit/s
over distances of 10 m to 100 m conducted by T-Nova GmbH, the University of
Ulm, Daimler Chrysler, the Fraunhofer Institute for Integrated Circuits Nuremberg
and the POF-AC Nürnberg.
Figure 2.25 shows theoretical considerations with respect to the POF bandwidth
at different launching angles (Gaussian shaped far field with 3 dB width calculated
relative to fiber NA) according to [Bun99a]. With short lengths and small launch
NAs the light remains concentrated in areas with small propagation angles. The
small differences in propagation time result in large bandwidths. After approx.
100 m of fiber equilibrium mode distribution is just about reached through mode
mixing, and the influence of the launch conditions gradually disappears. This be-
havior corresponds to a great degree to the measurement results described above.
1,000
B3 dB [MHz]
rel. launch NA
500 (NAfiber = 1)
0.5 1.5
200 0.7 1.7
1.0 2.0
100 1.2
50
20
length [m]
10
10 20 50 100 200 500
Fig. 2.135: Theoretical bandwidth with different launching conditions ([Bun99a])
The principle of peaking is demonstrated in Fig. 2.136 and 2.137 ([Zam00b],

[Ziem00a] and [Ziem00c]). A high pass filter which dampens lower frequencies
and lets high frequencies pass through without losses is switched between the
modulation input and the laser. We begin with an illustration of the electrical spec-
trum of the emitting signal at the laser with and without peaking (1.2 Gbit/s, NRZ,
pseudo-random bit sequence).
0 rel. power [dB]

without peaking
-10
with peaking
-20
-30
-40
-50
-60
-70 ¤ Giehmann
frequency [GHz]
-80
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Fig. 2.136: Modulation spectrum with and without peaking
In the experiment, the data rate was 1,200 Mbit/s with NRZ coding. The twin-
stage pre-distortion filter dampens the signal by 12 dB in the low frequency range
so that the higher frequencies can create a stronger modulation. For the pulses this
means steeper edges and overshoot at the beginning and end, hence the term
peaking, as shown in Fig. 2.137.
relative amplitude with peaking

18 without peaking
16
14
12
10
8
6
4
2 ¤ Giehmann 2000 time [ns]
0
0 10 20 30 40 50 60
Fig. 2.137: Laser modulation signal with/without peaking
The disadvantage of peaking can be clearly recognized in the diagram. The

peaks at the beginning and the end of the pulse must lie within the admissible ope-
rating range of the laser, i.e. between the threshold current and the maximum cur-
rent. This reduces the actual power per pulse compared with rectangular pulses.
Figure 2.138 summarizes the bit rates and transmitted distances of different
high-rate transmission systems using SI-POF ([Scha00], [Ziem00a], [Kich99] and
[Yas93], [Vin04a], [Vin04b] and [Ziem03g]). Chapter 6 contains detailed expla-
nations of the different systems. The diagram also shows the theoretical limits for
the bandwidth of standard NA-POF and DSI-POF (assuming NRZ coding and bit
rate = 2 u 3-dB bandwidth).
bit rate [Mbit/s] system with St.-NA

3,000 POF-AC system with DSI-POF
system with MC-POF
UMD-limit St.-NA
T-Nova UMD-limit DSI-POF
POF-AC
1,000
POF-AC
UNI Ulm
POF-AC
500
Bates ´93
Daimler
Chrysler IEEE 1394 Bates ´92
200
ATM 155
100BaseFX Kaiser ´92
100
10 20 50 100 200
POF-length [m]
Fig. 2.138: Bit rates of different POF systems (status 2003)
It is easily discernible that a number of systems with standard NA-POF are sig-
nificantly above the theoretical limits. Particularly for greater lengths the potential
for exceeding the limit is considerable. As shown in the next section, the practical
application presents some problems such as the bending behavior. It is generally
true that extreme dispersion compensation must be adaptive in its execution. That
means that above all the limit frequencies of the high passes must be adapted very
precisely to the frequency response of the link. If the frequency response changes,
the result will be too much or too little compensation so that the pulses become
distorted. Such a change may, for example, occur as a result of different lengths of
cable; however even a bend in the fiber may have the same effect. In commercial
systems it is desirable to avoid having to use automatic adaptations, such as is
necessary, for example, in 1000BaseT-systems, or having to provide specific
receivers for different cable lengths. One practical solution is to adjust the com-
pensation in such a way that there is just-tolerable over-compensation for short
lengths of fiber; based on this level of compensation, the next step is to select the
maximum fiber length for which this compensation is still just sufficient. A sche-
matic illustration is shown in Fig. 2.139.
increasing transmission distance
frequency response of the POF link (dispersion limited)
f f f f
compensation filter
(fixed)
f
resulting frequency response
f f f f
overcompen- optimized undercompen-
sation compensation sation
Fig. 2.139: Compensation of dispersion for various transmission distances
A proposal for increasing the bandwidth by direct interference with the optical
path is described in [Kal99]. By using a mode filter immediately after the trans-
mitter, the light angle range in the fiber is reduced as shown schematically in
Fig. 2.140. With this method it was possible to achieve an improvement in band-
width by 53% and 89% respectively for two standard NA-POF provided by
Mitsubishi and Toray. The losses of the mode mixer are approximately 2.5 dB
which is perfectly acceptable in many applications.
AN | 0.43 AN | 0.29
POF
source receiver
mode filter
Fig. 2.140: Increased bandwidth with a mode filter ([Kal99])
Basically, this method is equivalent to the method of light launching using a

smaller NA, though probably much easier to implement because no optical com-
ponents are required and only a simple mechanical clamp needs to be placed on
the fiber. If required, this can be repeated in the middle of the link or before the
receiver.
2.5.6 Bit Rates and Penalty
In general, optical transmission systems are set up in such a way that the system
bandwidth amounts to at least 50% of the bit rate with NRZ transmission. Thus,
500 MHz are needed to transmit 1000 Mbit/s. This means that the eye is complete-
ly open with ideally adapted filtering. In other words, the transition from the zero
symbol to the one symbol and vice versa takes place within the bit duration. If the
system bandwidth is smaller than half of the bit rate, then the symbol transition
needs more time resulting in a reduction of the vertical eye opening. This effect
must either be compensated for through adapted filtering or the reduced eye
opening is compensated for by a correspondingly higher receiving level. The
deterioration of the signal-to-noise ratio at the receiver through the bandwidth
limitation is called penalty (measured in dB). The relationship between signal-to-
noise ratio, receiving level and penalty is shown in Fig. 2.141.
U U
U1 U2
t t
system without noise and with system without noise and with
sufficiend bandwidth - the eye limited bandwidth - the eye in
open completely closed partially
penalty: 20·log(U2/U1)
U U
UN
US
t t
system with noise and with system with noise and with
sufficient bandwidth - the limited bandwidth - the
eye is open completely eye is more closed
SNR = 20·log(US/UN) SNR is decreased by penalty
Fig. 2.141: Definition of penalty

When describing the sensitivity of a receiver, measurements are always made at

the maximum bit rate. A possible penalty is always included. If there is sufficient-
ly large bandwidth, then only the noise should limit the sensitivity. The large
diode capacitance produces as a rule a relatively dramatic low-pass effect when
using large photodiodes, which are necessary for POF or PCS. The noise rises in
proportion to the signal with decreasing receiver impedance, i.e. one will accept
some penalty and work with as large an input resistance as possible.
Under laboratory conditions data communication can also be carried out with
high penalty. Modern bit error analyzers can transmit error free as long as the eye
opening amounts to some 10 mV. A typical eye diagram with high penalty is
shown in Fig. 2.142. Subsequently, the connection between system bandwidth and
penalty for a broad-band receiver at the POF-AC is illustrated.
Fig. 2.142: Data transmission with large penalty
In the example shown 820 Mbit/s were transmitted over 100 m of DSI-POF.
Although the eye was almost completely closed, an error free transmission was
possible. In a real system, however, certain detection would be relatively difficult
since the sampling moment and the decision threshold have to be re-adjusted very
exactly. Furthermore, there are no margin whatsoever for fluctuations in the laser
power or bending losses.
25 penalty [dB]
20 simulated
with fiber
15
10
0
0.10 0.20 0.30 0.40 0.60 0.80 1.00
bandwidth/bit rate [MHz/Mbit/s]
Fig. 2.143: Effect of system bandwidth on the penalty
2.6 Bending Properties of Optical fibers 143
The simulated values were determined by calculating the penalties with the aid
of PSpice analyses. A Gaussian-shaped filter was used as a low-pass system. The
measuring points were determined on a 20 m long standard POF with different bit
rates. The penalty was estimated from the eye diagram. The measured values
tallied greatly with the simulation down to 25% of the system bandwidth, e.g. a
transmission of 1 Gbit/s with a system bandwidth of 250 MHz. With higher bit
rates the penalty increases more quickly than in the simulation. One main reason is
that the frequency response only corresponds to a certain degree to idealized
Gaussian behavior. It hardly makes sense to use practical systems with more than
a 10 dB penalty.
The results show that an exact relationship does not have to necessarily exist
between the maximum bit rate and the fiber bandwidth. Furthermore, even with
bandwidth-limited systems relatively high data rates can be achieved under labora-
tory conditions if enough emitting power is available.
2.6 Bending Properties of Optical fibers
The sensitivity of optical fibers to bending is of special significance. In practical

applications installed links are never completely straight. Often they are fitted
around corners where 90º bends are a common occurrence. Even along a link in a
straight cable duct there are many small bends, for example, in places where
cables are hold with cable ties.
When being assembled, the fiber must also be able to withstand mechanically
tight bends. In many applications there is continuous bending during operation, for
example in drag chains or with a data cable in a car door. That is why one diffe-
rentiates between different bending loads:
¾Static bending: involves the assessment of how much light is loss in bends.
These losses are to be taken into account in the power budget of the system.
The bending loss is measured in dependence of the bending radius.
¾Minimum bending radius during assembly: only characterizes what bends the
fiber can tolerate for a short time without being mechanically destroyed.
¾Repeated bending: in certain applications fibers must be able to tolerate 105 to
some 106 bends without being mechanically destroyed.
¾Reel change bending: arises in particular in drag chains (see also Chapter 9).
The following results have either been taken from data sheets or come from
measurements made at the Deutsche Telekom and as of 2000 at the POF-AC
Nürnberg. Now as before no standards exist for measurements of bending attenu-
ation. We mostly used a long fiber sample stimulated with as large a NA as pos-
sible and all modes were detected with the aid of an integrating sphere. On the
other hand many manufacturers measure with a small NA whereby much better
values automatically come about because the outer modes are more strongly
radiated in the bends.
144 2.6 Bending Properties of Optical fibers
Nevertheless, the series of measurements cannot always be compared exactly.

In addition to the wavelength the bending attenuation can also depend to a great
extent on the primary coating material and on the lateral forces within the bend.
The coupling and detection conditions are also always included with short samp-
ling lengths.
2.6.1 Bending Losses in SI-POF
The essential parameters which determine the bending sensitivity of a fiber are the
diameter and the numerical aperture. The larger the NA, the narrower the permis-
sible bending radii may be in relation to the fiber diameter. Figures 2.144 and
2.145 show the losses for bends of different commercially available fibers accor-
ding to information in the data sheets ([Tor96a] and [Asa97]).
The Fig. 2.144 shows the bending losses of two different SI-POFs with some-
what different NAs. You can clearly see that larger NAs reduce the bending
losses.
bend losses [dB]

5.0
fiber
4.0
PFU-CD-1001
AN = 0.46
3.0
PGU-CD-1001
AN = 0.50
2.0
1.0
0,0
0 5 10 15 20 25 30 35 40
bend radius [mm]
Fig. 2.144: Loss for 360° bend according to [Tor96a]
Figure 2.145 shows losses resulting from bends in a standard NA-POF, a low-
NA-POF and a multi-core fiber (see Chapter 2.3).
The low-NA-POF shows significantly larger losses compared to a standard
NA-POF. Due to the smaller individual core diameters, the bending sensitivity of
the multi-core fiber is comparable with that of the standard NA-POF despite the
smaller NA.
If many bends directly follow each other, attenuation does not increase propor-
tionally with the number of bends because there is less and less energy present in
the higher mode groups. Figure 2.146 shows a measurement of the bending losses
for different POF according [Hen99].
6
bend losses [dB] fiber
5 TC 1000 (AN = 0.485)
4 NC 1000 (AN = 0.25)

NCM 1000 (AN = 0.25)
3
0 bend radius [mm]

0 10 20 30 40 50 60 70 80
Fig. 2.145: Loss for a 360° bend according to [Asa97]
10.0
loss [dB]
5.00 NC 1000 (Low-NA)
no longer available at
2.00 the market
1.00
0.50 AC 1000 (DSI) PFU 1000 (St.-NA)
0.20 MH 4000 (DSI-POF)

0.10
0.05
number of turns
0.02
0 1 2 3 4 5 6 7 8 9 10
Fig. 2.146: Bending loss depending on number of turns ([Hen99])
The measurements were taken at 650 nm with LED-launch and a mode mixer.
The bending radius was 32 mm and the bends were located at the beginning of a
50 m sample length.
PFU 1000 is a standard NA-POF, while MH 4000 and AC 1000 are double-step
index POF. Their losses are approximately identical and up to 10 windings are
significantly below 1.0 dB. By comparison, the low-NA-POF NC 1000 is in the
range of 10 dB, which is too much for deployment in practical applications. The
ATM forum stipulates an admissible bending radius of 25 mm and at this radius
the attenuation was already above the range of measurement. Meanwhile, DSI-
POF offer significantly improved bending characteristics at comparable NA.
Figures 2.147 and 2.148 demonstrate the losses over the inverse bending radius
and the number of windings for a (genuine) low-NA-POF (NC 1000) and a stan-
dard NA-POF [Hen99].
12
loss [dB] 12 mm
10 10 turns
8 turns
8
15 mm
6 turns
6
4 turns
4
21 mm 2 turns
26 mm
2 32 mm
39 mm
0
0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09
inverse bend radius [mm-1]
Fig. 2.147: Bending loss of a PFU-CD-1000 ([Hen99])
Under UMD conditions, the bending losses should increase proportionally to

the inverse bending radius. In practice, however, this only takes place below a
bending radius of around 20 mm. It appears that the real equilibrium mode distri-
bution reduces the losses above a certain radius. The reason for this is the smaller
weighting of modes that have a large propagation angle, which, as already men-
tioned, are particularly sensitive to losses at bends.
Basically the low-NA-POF in Fig. 2.148 shows the same behavior, albeit for
significantly greater radii and due to the smaller NA.
16
loss [dB] 28 mm 10 turns
14
12 8 turns
32 mm
10
6 turns
8
4 turns
6
4 39 mm
2 turns
50 mm
2
0
0.020 0.022 0.024 0.026 0.028 0.03 0.032 0.034 0.036
inverse bend radius [mm-1]
Fig. 2.148: Bending loss of a NC-1000 ([Hen99])

2.6.2 Bending Losses in GI Fibers
For graded index POF slightly different conditions apply for bending sensitivity
compared with step index profile fibers. Here it is not the total reflection at the
core-cladding interface but the continuous bending in the index profile that is res-
ponsible for light guiding. In addition, there is a fundamentally different distri-
bution in the near and far field. Figure 2.149 shows a measurement for GI-POF
according to [Ish95].
bend losses [dB]

20
GI-POF, doped by:
10 MMA/DPS, AN = 0.29
5 MMA/BBP, AN = 0.21
2
1
0.5
bend radius [mm]
0.2
0 10 15 20 25 30 35 40 45 50 55
Fig. 2.149: Loss of two GI-POF ([Ish95]) for a 90° bend
Due to the different dopants used, the two samples with a core diameter of
0.5 mm each have a different NA, which has a very significant effect on the ben-
ding losses. Despite the smaller core diameter the losses for a 25 mm bend are still
significantly higher than the values for a SI-POF or a DSI-POF. Here, too, a
reduction in the core diameters leads to lower bending losses.
[Aru05] describes how the bending losses in PMMA GI POF can be signifi-
cantly reduced. In addition to an optimized index profile an additional PVDF layer
(polyvinylidenfluoride) was applied to the core with parabolic profile resulting in
a semi-GI-POF which combines high bandwidth with low bending losses. The
losses of a 90° bend are shown later in Fig. 2.205 compared with a conventional
PMMA GI-POF. (The sample length was 100 m.) Even with a bending radius of
5 mm there was no measurable increase in attenuation. The different methods for
reducing bending losses in PMMA GI-POF and PF-GI-POF are described in
greater detail in Section 2.8 on fiber production. Examples of measurements are
also shown.
2.6.3 Change of Bandwidth by Bends
However, bends do not only contribute to additional losses, but also have an effect
on bandwidth because certain mode groups are selectively attenuated. This effect
is exploited in mode filters and mixers.
Figure 2.150 (according to [Rit93]) shows what the effect of a 720º bend at the
beginning of a 50 m long POF link has on the measured bandwidth. In this case
the light is launched with AN = 0.10.
bandwidth [MHz]
160
ØPOF:
140
250 μm
120 500 μm
750 μm
100 1000 μm
80
60
0.00 0.02 0.04 0.06 0.08 0.10 0.12

rel. inverse bend radius [ØPOF-1]
Fig. 2.150: Change of bandwidth by bending the fiber according to [Rit93]
Due to the low launch NA, the bandwidth is relatively large (80 MHz · 100 m).
In the case of tight bending radii at the beginning of the fiber there is mode mixing
so that the bandwidth is significantly reduced sometimes. This effect is naturally
more pronounced for smaller diameters. In the illustration selected here above the
inverse relative bending radius, relative to the core diameter, the effect of the core
diameter should disappear. It seams to be, that the effect described above of the
larger bandwidth for thinner fibers is already dominant here due to the more mode
dependent processes.
Comprehensive investigations of the effect of bends on the bandwidth of POF
links were presented in [Mar00]. The test fiber consisted of a 100 m long standard
NA-POF; 360º bends were inserted at the beginning of the link, after 25 m, after
50 m, after 75 m or at the end of the link. The source consisted of a 655 nm laser
diode, the NA of which could be adapted through different optics from 0.10 to
0.65. The bandwidth and the attenuation of the overall link were measured without
bends and with bending radii of 6.4 mm, 11.1 mm and 13.8 mm. The results are
shown in Fig. 2.151.
When light is launched into the fiber using a large NA, the original bandwidth
of approximately 33 MHz can be increased significantly. However, large improve-
ments with small bending radii occur at the expense of large additional losses. The
biggest gain in bandwidth is obtained with a bend in the middle because this
means that many modes of the first 50 m are filtered out and EMD is not com-
pletely regained in the remaining 50 m. The changes in attenuation are largely
independent of the length since the mode field is well filled out everywhere.
launch NA: AN = 0.65 launch NA: AN = 0.10

bandwidth over 100 m [MHz]
100
90
80
70
60
50
40
30
-1
-2
-3
-4
-5
0 25 50 75 100 0 25 50 75 100
loss
[dB] bend position [m] bend position [m]
radius 6.4 mm radius 11.1 mm radius 13.8 mm

bandwidth without bendings
Fig. 2.151: Influence of a bend to bandwidth and attenuation ([Mar00])
When light is launched into the fiber using smaller NA, the relative gain in
bandwidth compared to the original - approximately - 60 MHz is not as big.
Therefore the optimum position for the bend is clearly nearer to the end since the
mode field must first be filled. Again, tight radii have more effect. The additional
attenuation increases significantly when the bends are moved to the end, since at
the beginning of the fiber there are hardly any higher mode families in existence.
These results also confirm clearly for the existing assumptions with respect to
mode propagation in a coupling length of some 10 m.
2.6.4 Bends on PCS, Multicore Fibers and thin POF
A very simple method to decrease bending radii is to reduce the core diameter
while otherwise retaining identical parameters. If you wish to maintain the
advantage of the simple handling of ready-made thick fibers, then there is the
possibility of fiber bundles or multicore fibers respectively.
Fig. 2.152 and 2.153 show the measured bending losses, each with a bend of
360° in the middle of the sample, with UMD launch and measured with an inte-
grating sphere. A 10 m long fiber was used for the MC-GOF. The range of the
bending radii lay between 2 mm and 100 mm. The bending attenuation measured
lies below 0.1 dB.
0.06
bending loss [dB]
0.05
0.04
0.03
0.02
one bend
0.01 by 360°
0.00
inverse bending radius [mm-1]
0.0 0.1 0.2 0.3 0.4 0.5 0.6
Fig. 2.152: Bending losses of MC-GOF, Schott
The bending losses of the MC-POF were measured on a 100 m long sample in
order to guarantee as much mode equilibrium as possible. The bend (360°) was
made in the middle of the fiber length. Due to the different relations between
mode coupling and absorption the EMD conditions for 520 nm and 650 nm only
differ slightly. That is the reason for the somewhat different bending losses.
bending loss [dB]

1.00 100 m fiber length
at 650 nm
at 520 nm
0.10
bending radius [mm]

0.01
0 5 10 15 20 25 30
Fig. 2.153: Bending losses of MC-POF, 37 cores, 1 mm total diameter
In many areas the 200 μm PCS is used because it permits smaller bending radii.
Fig. 2.154 illustrates quite graphically that the same physical characteristics are
also valid for these fibers. Here the bending losses are given versus the relative
bending radius in relation to the fiber diameter. The numbers in brackets indicate
the bending radius in millimeters for the PCS. Both fibers thus have in relative
terms an identical bending sensitivity.
bending loss [dB]

4.0
3.5 200 μm PCS
3.0 1 mm POF
2.5
2.0
1.5
1.0
0.5 bending radius [u Kern]
0.0
0 5 10 15 20 25 30 35 40 45 50
(0) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Fig. 2.154: Bending losses of PCS and POF in comparison
Thin POF could be used as an alternative to PCS in many areas when tight
bending radii are indeed required, but the attenuation and the temperature range of
the POF are satisfactory. A comparison between a 250 μm SI-POF and a 200 μm
PCS, measured at 650 nm with full launch for 5 m long samples, is illustrated in
Fig. 2.155.
10.0
bending loss [dB]
3.0
1.0 250 μm POF
0.3
0.1
200 μm PCS
0.03
bending radius [mm]
0.01
0 1 2 3 4 5 6 7 8 9 10
Fig. 2.155: Bending losses of small diameter POF and PCS in comparison
The somewhat thicker POF also has somewhat higher bending losses. A tenth
of a dB is attained for the POF at a bending radius of 8 mm and 6 mm for PCS.
The bending losses of three different SI-POFs with different NAs are compared
in Fig. 2.156. The lowest losses are shown by the 300 μm thick POF with a high
NA. The 250 μm and 500 μm thick POFs have almost identical bending attenu-
ations. It is thus indicative that the NA is by far the most important factor for the
bending losses. Consequently, you should always choose fibers with the largest
possible NA for particularly tight radii, unless you decide to go back to multicore
fibers. In addition, the latter have the advantage of offering an even greater
bandwidth.
10.00
bending losses [dB] 250 μm POF (AN = 0.63)
500 μm POF (AN = 0.50)
1.00 300 μm POF (AN = 0.63)
0.10
0.01
bending radius [mm]

0.001
0 1 2 3 4 5 6 7 8 9 10 11 12 13
Fig. 2.156: Comparison of bending losses of various SI-POF (different NA)
More recent measurements of bending losses of four different SI fibers, each

with cladding and made available from Toray Germany, are shown in Fig. 2.157.
Fibers with a large NA (0.63) were used for this measurement. They allow
considerably smaller bending radii without remarkably decreasing the attenuation
and bandwidth.
3.0
D [dB] 7.5 u r
7.5 u r
7.3 u r
1.0 8.0 u r
0.3
0.1
250 μm 750 μm
1000 μm
0.03
500 μm
0.01 r [mm]
0.3 1.0 3.0 10.0 30.0
Fig. 2.157: Comparison of bending losses of various SI-POF
Bending radii are drawn in the picture with which a bend (360°) results in
exactly 1 dB additional attenuation. With the four fibers with their 250 μm to
1000 μm core diameters this is the case each with a seven-fold to eight-fold fiber
radius, i.e. a bending radius between 0.9 mm and 4 mm. As a comparison, the
bending losses of a 125 μm SI-POF ([Witt04]) are shown in Fig. 2.158.
bending losses [dB]

2.0
1.0
0.5
0.2
bending radius [mm]
0.1
0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
Fig. 2.158: Bend losses of a 125 μm SI-POF ([Witt04])
Optimedia has made available samples of a thinner PMMA GI-POF. The ben-
ding losses of this fiber with overfilled launch (LED) and a launch with a laser
(AN = 0.10) are shown in Fig. 2.159. Both measurements were carried out at
650 nm with a 5 m long fiber.
bending losses [dB]

5.0
OM-Giga
500 μm/750 μm
2.0 360°-bends
overfilled (LED) O = 650 nm
1.0 5 m fiber
0.5
laser launch
0.2
0.1
0 20 40 60 80 100
bending radius [mm]
Fig. 2.159: Bending losses of a 500 μm PMMA GI-POF
A bending radius of 15 mm is still not a problem for collimated light whereas a

high bending attenuation arises below a bending radius of 30 mm with an LED
launch. You could argue, of course, that laser sources should always be used for
GI fibers in order to utilize the high bandwidth. Nevertheless, it is imperative that
manufacturers reduce the bending losses.
Finally, some results from a project work [Bau06] are shown. First, Fig. 2.160
compares the bending losses of three Toray fibers with different core diameters:
500 μm, 750 μm and 1,000 μm. The NA of the three fibers is the same. As expec-
ted, the bending radius for a given attenuation is reduced nearly proportional to the
fiber diameter. Only with very thin fibers does the effect of stronger mode-depen-
dent attenuation make itself noticeable.
10
bending losses [dB] fiber type: Toray PFU
AN = 0.47
measured with 650 nm LED
1 bend 360°, 10 m fiber
1
0.1 1,000 μm
750 μm
500 μm
0.01
bending radius [mm]

0.001
0 10 20 30 40 50
Fig. 2.160: Bending losses of various standard-NA-POF
The bending losses of 1 mm POF from three manufacturers are compared in

Fig. 2.161. Since the NAs of the fibers are not exactly equal, the bending attenu-
ations differ somewhat. In practice, however, these small deviations should hardly
play a role.
10
bending losses [dB] fiber type: St.-NA
measured with 650 nm LED
1 bend 360°, 10 m fiber
1
0.1
0.01
0 10 20 30 40 50
bending radius [mm]
Fig. 2.161: Bending losses of various standard-NA-POF
2.7 Materials for POF 155
2.7 Materials used for POF
2.7.1 PMMA
The material most frequently used for polymer fibers is the thermoplastics PMMA
(Polymethylmethacrylate), better known as Plexiglas. Figure 2.162 shows the
structure of the monomer and its polymer chain.
MMA PMMA
H
H C H CH3 CH3 CH3 CH3
H
C C C CH2 C CH2 C CH2 C CH2
C H C C C C
O H O
O C OCH3 O OCH3 O OCH3 O OCH3
H H
Fig. 2.162: Molecular structure of PMMA
PMMA is produced from ethylene, hydrocyanic acid and methyl alcohol. It is

resistant to water, lyes, diluted acids, petrol, mineral oil and turpentine oil. PMMA
is an organic compound forming long chains with typical molecule weights
around 105. Essential from the point of view of optical transparency of the
material is the amorphous structure of the polymerized material. The density of
PMMA is 1.18 g/cm3. Its tensile strength is approximately 7-8 kN/cm2 ([SNS52]).
The refractive index of PMMA is 1.492 and the glass transition temperature Tg
lies between +95°C and +125°C. At room temperature and 50% relative humidity
the material can absorb up to 1.5% water, which also affects the attenuation
characteristics.
Table 2.17 presents further properties of PMMA:
Table 2.17: Properties of PMMA (typical values)

Parameter Unit Value
refractive index - 1.492
glass transition temperature Tg °C 115
density g/cm³ 1.18
absorption of water up to saturation % 0.5
thermal conductivity: W/mK 0.17
thermal heat expansion coefficient: mm/mK 0.07
Rockwell hardness (M), - 95
Shore hardness (D) 70
tensile strength N/mm² 76
resistivity Ohmcm 1015
breakdown strength kV/mm 20 - 25
spontaneous combustion temperature °C approx. 430
156 2.7 Materials for POF
As can be seen in the illustration, each MMA monomer has a total of eight C-H
bonds. The vibrations of this compound, or more precisely its harmonic waves are
a main cause for the losses encountered in PMMA polymer fibers. The attenuation
resulting from absorption at the respective wavelength is shown in [Mur96] and
[Koi96c] (see Fig. 2.163 and table 2.18). In particular the harmonic waves at
627 nm (6th harmonic wave) and 736 nm (5th harmonic wave) essentially deter-
mine the level of attenuation within the application range of PMMA-POF because
these are not narrow absorption lines but relatively wide bands. Further causes for
attenuation will be discussed in the chapter titled Characteristics.
108
attenuation [dB/km]
106 molecule
104 C-H
102 C-D
C-F
100
C - Cl
10-2
10-4
10-6
10-8
500 1000 1500 2000
wavelength [nm]
Fig. 2.163: Absorption lines of C-X-bounds according to [Gra99] and [Mur96]
Quite early in the history of this technology, the idea came up to reduce the ab-
sorption losses of polymer fibers by using different materials in which less or no
C-H bonds were present. However, it is not easy to eliminate these; instead, the
hydrogen atoms are replaced by other atoms of the 7th main group. A heavier core
will result in a lower vibration frequency, thus moving the attenuation bands to a
larger wavelength. The illustration shows the attenuation bands for deuterium
(heavy hydrogen with the atomic weight 2), fluorine (atomic weight 19) and
chlorine (atomic weight 35 or 37, see also [Bau94]). Generally, the materials for
polymer fibers can be divided into three groups:
¾compounds containing hydrogen
¾compounds with partial substitution of hydrogen
¾compounds with complete substitution of hydrogen
Table 2.18: Absorption bands position of carbon bonds ([Gra99])

Oscillation C-H C-D C-F C-Cl C=0 O-H
O [nm] O [nm] O [nm] O [nm] O [nm] O [nm]
Q0 3,390 4,484 8,000 12,987 5,417 2,818
Q1 1,729 2,276 4,016 6,533 2,727 1,438
Q2 1,176 1,541 2,688 4,318 1,830 979
Q3 901 1,174 2,024 3,306 1,382 750
Q4 736 954 1,626 2,661 1,113 613
Q5 627 808 1,361 2,231 934 523
Q6 549 704 1,171 1,924 806
Q7 626 1,029 1,694 710
Q8 566 919 1,515 635
Q9 830 1,372
2.7.2 POF for Higher Temperatures
Fibers with high resistance to heat are especially needed for use in certain areas of
automotive engineering (engine compartment) and automation technology. In the
passenger compartment of a vehicle a maximum of +85°C will arise. PMMA-POF
can easily be used with such temperatures. In the area near the center console or
under the roof temperatures can also go up to over +100°C and near the engine to
+125°C. Summaries of the data published so far and of comprehensive investi-
gations at the POF-AC Nürnberg can also be found in [Poi03a] and [Poi03b]. On
the whole the following methods for increasing the resistance to heat of polymer
fibers have been presented:
¾Cross-linking of PMMA: cross-linking between polymer chains can be
generated by chemical effects or by UV irradiation which results in a rise of
Tg. At the same time, however, the scattering and the mechanical charac-
teristics become worse.
¾Polycarbonate: PC has a considerably greater Tg compared with PMMA and
is likewise transparent. Fibers made of this material have been produced on a
large scale. PC fibers, however, age relatively quickly in combination with
humidity.
¾Elastomers: fibers made of this material could be used up to +170°C and
show very low attenuation. So far, they have only been produced as labora-
tory samples.
¾Alternative polymers: a series of other polymers such as cyclical polyolefins
have Tg up to +200°C.
When determining the thermal stability, a maximum increase in the kilometric
attenuation is established over a maximum period of aging. In case the aging pro-
cedures are thermally activated, then the permissible operating period decreases
almost logarithmically to the temperature. An example of the behavior of a stan-
dard PMMA-POF can be seen in Fig. 2.164 (measurements were made at the
POF-AC). The increase in losses is represented here vs. the temperature. With an
approx. 10 K increase in temperature the speed of ageing increases about one
order of magnitude.
1000
increase of
attenuation
coefficient
dB/(km1000 h)
100
520 nm
590 nm
10
650 nm
temperature [°C]
1
70 75 80 85 90 95
Fig. 2.164: PMMA-POF ageing
2.7.2.1 Cross-Linked PMMA

One of the most obvious methods for more heat-resistant POFs is the use of cross-
linked PMMA, generally referred to as modified PMMA. The attenuation curves
of such fibers are summarized in Fig. 2.165. The fibers of the PHK Series are sold
by Toray ([Tor96a] and [LC00a]). Important parameters are:
¾Core/cladding: PMMA/fluoropolymer
¾Diameter. 0.5 mm, 0.75 mm. 1.0 mm and 1.5 mm
¾NA/aperture angle: 0.54/65°
¾Lowest attenuation at 650 nm (for 1 mm): <300 dB/km
¾Permissible bending radius: 9 mm
¾Operating temperature: -40°C to +115°C
¾Available as single fiber and bundle with 18 fibers à 0.5 mm
The attenuation measurement at the POF-AC is shown in the figure. A first
version of the Toray fiber has already been presented [Tan94a]. Another fiber
manufacturer active in this field is Asahi Chemical. A first sample was introduced
in 2003 under the designation H-POF and the measurement results were presented
in [Poi03a]. The manufacturer provides the following data:
¾Core/cladding: cross-linked PMMA/P-FEP
¾Primary coating: ETFE (black Tefzel®)
¾Numerical aperture (after 2 m): 0.65
¾Core diameter: 1.0 mm
¾Primary coating: 1.51 mm/2.3 mm (MOST specification)
¾Attenuation: 540 dB/km (measured at 657 nm)

¾Minimum bending radius: 5 mm
¾Bandwidth: 30 MHz · 100 m
Samples of PMMA-POF with different degrees of cross-linking were produced
in 2002 to 2004 by the RPC Institute in Tver near Moscow. The measurement
results of a sample are shown in the picture. The maximum application tempe-
rature lies at +130°C with an attenuation of about 800 dB/km at 650 nm.
4.000
attenuation [dB/km]
2.000
1.000
500
Tver-POF1
H-POF
200
[Tan94a]
PHKS wavelength [nm]
100
400 500 600 700 800
Fig. 2.165: Attenuation of cross-linked PMMA-POF
On the whole it is true for this type of fiber that a higher degree of cross-linking
leads to higher application temperatures, whereby the scattering is also greater so
that the losses increase. A short piece of a Tver fiber sample exposed to red light
is shown in Fig. 2.166. The high degree of scattering leads to a clearly visible
lateral emission.
Fig. 2.166: Cross-linked PMMA-POF (sample from Tver)

2.7.2.2 Polycarbonate POF

The first polymer fibers on the basis of polycarbonate were introduced in 1986 by
Fujitsu ([Ish92b] and [Koi95]). The attenuation lay at 800 dB/km at 660 nm and
450 dB/km at 770 nm respectively. The maximum operating temperature was
given at +130°C. Similar data were published in [Min94] - see also Fig. 2.167.
In 1992 [Tesh92], Asahi introduced another PC-POF called Luminous H. With
an application temperature of +125°C the attenuation was 600 dB/km at 660 nm.
The fiber NA was 0.78 and the bandwidth 17 MHz · 100 m. The relatively large
NA of most PC-POFs can be explained by the high refractive index of PC which
amounts to about 1.59. If PMMA is used with n = 1.49 as cladding material, the
result is then AN = 0.55.
Mitsubishi sells a type of fiber, the ESKA FH4001-TM, with a temperature
capability up to +125°C. The specific parameters for this type are:
¾Application temperature range: -55°C to +125°C
¾Application temperature at high humidity: +85°C
¾Maximum attenuation at 770 nm: 800 dB/km
¾Core/cladding material: polycarbonate/fluoropolymer
¾Refractive index core/cladding: 1.582/1.392
¾Numerical aperture: 0.75 ± 0.01
¾Core/ cladding diameter: 910 ± 50 μm / 1000 ± 60 μm
¾Primary coating: 2.2 mm polyolefin elastomer
Laser Components GmbH offered another PC fiber. The last PC-POF shown in
Fig. 2.167 was introduced by Furukawa ([Hatt98], [Nish98] and [Irie94]),
whereby a material was used in which hydrogen atoms were partially replaced by
fluorine.
10,000 attenuation [dB/km]

Furukawa
Minami 1994
5,000 Laser Comp.
Mitsubishi
2,000
1,000
500 wavelength [nm]

400 500 600 700 800 900
Fig. 2.167: Various PC-POF

The fibers produced by Furukawa with a core diameter of 0.5 mm had a NA of

0.35 and 0.53 (elastomer as cladding material). The low-NA version attains a
bandwidth of 200 MHz 100 m. A data rate of 156 Mbit/s could be transmitted
over 80 m of fiber (200 Mbit/s over 70 m).
No change in length could be ascertained in ageing tests over 10 days at tempe-
ratures of +100°C to +155°C (Fig. 2.168) which corresponds to an improvement
by 20 K over conventional PC-POF.
10 length variation [%]
4
PC(AF)
2 PC-A
0
105 115 125 135 145 155
temperature [°C]
Fig. 2.168: Temperature resistance of PC ([Hatt98])
The different PC-POF from Furukawa are summarized once again in

Fig. 2.169. Unfortunately, we do not know of any other work carried out by this
company.
5,000
attenuation [dB/km]
PC(AF)
2,000 [Hatt98]
D-POF
[Irie94]
1,000
PC-POF
[Irie94]
500
wavelength [nm]
300
400 500 600 700 800 900
Fig. 2.169: Data by Furukawa 1994-1998 (Polycarbonate)
The greatest disadvantage of PC-POF is its poor stability in regard to humid

heat. BAM tests on the aging of different POFs are summarized in Fig. 2.170.
What is surprising here is that PC-POF broke down before standard PMMA-POF.
120%
relative transmission at 92°C / 95 % RH
wavelength: 650 nm;
100%
sample length: 10 m
SI-mod. PMMA
80%
60%
40% SI-PMMA
20% SI-PC
ageing time [h]

0%
0 500 1,000 1,500 2,000 2,500 3,000 3,500
Fig. 2.170: Ageing behavior of various POF ([Daum03c])
2.7.2.3 Elastomer POF

Possibly the most suitable material group for heat-resistant POF are elastomers. A
number of institutes have already produced samples, but real product development
is still missing.
10,000
attenuation [dB/km]
[Ish92]
5,000 HPOF-S
HPOF-Sb
[Suk94]
[Zei03]
2,000
1,000
400450 500 550 600 650 700 750 800 850 900
wavelength [nm]
Fig. 2.171: Attenuation of various EOF
The attenuation curves of different EOFs (elastomer optical fiber) are compared
in Fig. 2.171. Particulars of the following fibers are compared:
¾Elastomer POF, produced by G. Zeidler (see [Zei03])

¾HPOF-S (Hitachi), data sheet information
¾HPOF-Sb (Hitachi), POF-AC measurements (1.5 mm core diameter)
¾2 mm silicone elastomer POF, AN = 0.54 [Ish92b]
¾POF made of ARTONTM, cyclical olefin, [Suk94]
You can clearly see that the attenuation spectra are quite similar to those of PC
fibers. The lowest losses lie in the range around 500 dB/km which is entirely
acceptable for use in vehicle networks or for parallel connections.
Typical representative fibers for both EOF and PC POF are compared in
Fig. 2.172. The similar functional groups lead to only slight differences in the loss
spectra.
3000 attenuation [dB/km]

silicone-POF
2000
PC-POF
1000
500
500 550 600 650 700 750 800 850
wavelength [nm]
Fig. 2.172: PC-POF in comparison with silicone-POF
The most recent development by Asahi is particularly interesting. The HPOF-S

possesses the following specified parameters (see [Poi03b]).
¾Core/cladding material: elastomer/P-FEP
¾Primary coating: ETFE (black Tefzel®)
¾Numerical aperture (after 2 m): 0.65
¾Core/cladding diameter: 1.00 mm / 1.50 mm
¾Primary coating: 2.3 mm
¾Attenuation: 800 dB/km (measured at 660 nm)
¾Bandwidth: 250 MHz · 100 m
Since the cladding could not be extruded directly, it was produced as a tube. At
250 μm it is relatively thick. Practical technical production methods for such
fibers are undoubtedly possible.
When aged under high temperatures, the attenuation of this fiber even dropped
to values around 300 dB/km. Whether this was due to the possible drying of the
fiber or by improving the adhesion of the cladding on the core could not be
determined. The material system thus shows some enormous potential.
2.7.2.4 Cyclic Polyolefines

A theoretically useful group of materials for POF are also the polyolefins.
Figure 2.173 shows a possible structure. These materials can also be produced
transparent. Low losses are theoretically possible because of their amorphous
structures.
H H H H
C C C C
H R
x
R’ R’ y
Fig. 2.173: Molecule structure COC
Some principal characteristics of such materials are:

¾Low water absorption
¾Theoretically more transparent than PMMA
¾Refractive index n = 1.56, makes another range of NA possible as well as the
production of different index profiles
¾Tg typically > 150°C
Manufacturers of such polymers are among others Ticona and JSR. It is not
foreseeable when test fibers made from this very promising material system will
be produced again.
2.7.2.5 Comparison of High-Temperature POF

So far the following temperature-resistant fibers have been summarily described:
¾Cross-linked PMMA (>130°C)
¾Polycarbonate (115°C)
¾Partially fluorinated polycarbonate (145°C)
¾Silicone elastomers (>150°C)
¾Thermoplastic resins (145°C)
¾ARTONTM (Fujitsu) (170°C)
The data of these different fibers have been compiled in Tables 2.19 to 2.21.
*)
temporary data sheet, the fiber is not presently available
**)
modified PC - partially fluorinated according to the authors’ information
***)
different data on materials, but with identical attenuation curves
Table 2.19: Polycarbonate-POF

Parameter Mitsubishi Producer Furukawa Furukawa**) Laser
FH 4001- B*) [Irie94] [Hatt98] Comp.
TM [Hatt98] [Nish98]
core diameter 910r50 μm 940r20 μm 910 μm 500 μm 1 mm
cladding thickness 40-50 μm 30 μm n. a. n. a. n. a.
NA 0.75 0.54 n. a. 0.30 0.61
x dB/km @ y nm 800@770 2000@633 400@660 460@650 800@770
1500@780 700@760 300@780
bandwidth n. a. n. a. n. a. 200MHz100m n. a.
max. temperature +125°C +125°C +125°C +145°C n. a.
***)
core material n = 1.582 n = 1.586 PC(A) D-POF, PC-AF PC
cladding material n = 1.392 n = 1.491 n. a. n. a. n. a.
Table 2.20: Properties of modified PMMA-POF

Parameter Toray Hitachi Tver-POF Toray
PHKS- H-POF (Sample [Tan94a]
CD1001-22 2002)
principle mod. PMMA cross linked PMMA Copolymer Copolymer
core diameter n. a. 1 mm 1 mm 1 mm
cladding thickness n. a. 250 μm 30 μm n. a.
NA 0.54 0.65 >0.50 n. a.
x dB/km @ y nm 300@650 540@660 800@660 250@650
bandwidth n. a. 30 MHz100m n. a. n. a.
max. temperature +115°C +130°C +130°C Tg = 135°C
core material PMMA PMMA PMMA Copolymer
cladding material n. a. P-FEP n. a. n. a.
jacket PP ETFE n. a. n. a.
Table 2.21: Properties of different high temperature-POF

Parameter Hitachi Hitachi Bridgestone Zeidler Fujitsu
HPOF-S [Sas88] [Ish92b] [Zei03] [Suk94]
principle silicone resin silicone elastomer elastomer
core diameter 1.0 mm 1 mm n. a. 1 mm 1 mm
cladding thickness 0.25 mm 0.5 mm n. a. n. a. n. a.
NA 0.65 0.62 0.54 0.44/0.25 n. a.
x dB/km @ y nm 800@660 660@650 700@660 800@770 800@680
900@780 450@770
bandwidth 25 MHzkm n. a. n. a. n. a. n. a.
max. temp. n. a. >150°C +150°C +150°C Tg = 171°C
core material n. a. ester based silicone elastomer ARTON
thermosetting resin
cladding P-FEP ethylen tetrafluoride - n. a. elastomer n. a.
propylene hexafluo-
material ride copolymer
fluorcopol.
jacket Tefzel (ETFE) n. a. n. a. without n. a.
Fig. 2.174 shows an aging experiment at +130°C with different fibers described
above. The most suitable ones at these temperatures were evidently the EOF and
the PC-POF.
10,000
attenuation [dB/km]
5,000
T = +130°C
TVER 2002
2,000
1,000 FH 4001
500 PHKS
200 HPOF-S
measuring time [hours]
100
0 100 200 300 400 500 600 700
Fig. 2.174: Ageing of various POF at high temperatures
The PC-POF from Mitsubishi (FH4001) only shows a moderate increase while
the two POFs made of cross-linked PMMA aged more quickly. The EOF even
gets better during the measurement period. Particularly noticeable is the clear drop
in attenuation after 15 hours. This was the point at which the temperature was
raised in the climate test chamber. It was noticeable that the bandwidth of the EOF
had dramatically diminished after this treatment. The combination of both events
provides the explanation that the adhesion of the cladding onto the core was
clearly improved by the high temperature so that even higher modes can now be
guided.
2.7.3 Polystyrene-Polymer Fibers
Another candidate for the production of polymer optical fibers is polystyrene (PS),
the molecular structure of which is shown in Fig. 2.175 ([Ram99]).
H H H H
C C C C
C H C H
H C C H H C C H
H C C H C C
C H H
C
H H n
Fig. 2.175: Molecule structure of PS

Theoretically, the attenuation of PS is partly below that of PMMA, as the

following theoretical estimate of losses in [Kai89] shows - without taking into
account propagation effects and the effects of claddings (see table 2.22).
Table 2.22: Theoretical attenuation of different polymers according to [Kai89]
Material Wavelength Rayleigh- UV-Ab- C-H- Sum Total

Scattering sorption Absorption
PMMA 520 nm 28 dB/km 0 dB/km 1 dB/km 29 dB/km
570 nm 20 dB/km 0 dB/km 7 dB/km 27 dB/km
PS 552 nm 95 dB/km 22 dB/km 0 dB/km 117 dB/km
PMMA- 680 nm 10 dB/km 0 dB/km 0 dB/km 10 dB/km
d8 780 nm 6 dB/km 0 dB/km 9 dB/km 15 dB/km
To date, PS-POF have been manufactured e.g. by Toray (first PS-POF 1972),
NTT (1982) and CIS in Tver (1993). The initial fibers had an attenuation of over
1,000 dB/km; later on it was possible to reduce this to 114 dB/km at 670 nm
([Koi95]). The NA of these fibers which can be used at temperatures up to 70°C is
0.56, i.e. a little higher than that for the standard PMMA-POF. Figure 2.216 shows
the attenuation behavior of a PS-POF ([Ram99], red curve and [Zub001b]).
1000
attenuation [dB/km]
800
600
400
200
[Zub01b]
wavelength [nm]
100
500 550 600 650 700 750 800 850
Fig. 2.176: Attenuation spectrum of PS-POF acc. to [Ram99] and [Zub01b]

The refractive index of PS is n = 1.59 so that it is possible to use PMMA for the
optical cladding (n = 1.49), as is possible for PC (n = 1.58). The glass transition
temperature of PS is approx. 100°C and therefore approx. 5 K lower than that of
PMMA. Hitherto there has been no reason to replace the PMMA-POF by PS so
that this material is not of any practical significance.
2.7.4 Deuterated Polymers
As has been illustrated in Fig. 2.163, a significant reduction in the absorption

losses of polymers can only be achieved by substituting the hydrogen with heavy
atoms. This would seem to be achieved most simply by replacing it with deu-
terium. This isotope has twice the atomic mass compared to hydrogen. In nature,
approximately 0.0156% of all hydrogen atoms are deuterium (1 atom in every
6,400). Chemically, deuterium behaves the same way as hydrogen so that it
simply makes sense to use so-called heavy water (D2O) as a base material for this
synthesis. Table 2.23 shows data of different POF based on deuterated polymers.
Table 2.23: Data of deuterated materials

Ref. Year Producer Attenuation at: Remarks
dB/km nm
[Koi95] 1977 Du Pont 180 790 first deuterated SI-POF
[Koi96c] 1982 NTT 20 680 SI-POF
[Lev93] 1993 CIS 120 650 core: 200-1000 μm,
180 850 AN = 0.48, to 70°C
[Koi92] 1993 Keio Univ. 56 688 core: 500 μm, MMA-BBP-d8,
[Khoe94] 94 780 2.000 MHz km
[Kon02] 2002 Keio Univ. 58 650 g = 3.4; 511 MHz 300 m
109 780 Tg = 105°C
127 580
According to [Koi95], the first deuterated SI-POF was produced by DuPont in

1977. In 1982, NTT ([Koi96c]) produced a SI-POF in deuterated material with a
minimum attenuation of 20 dB/km at 680 nm. It was not until the year 2000 that
this attenuation value was improved with the introduction of LucinaTM-POF.
Figure 2.177 shows further attenuation curves for POF made with deuterated poly-
mers; all examples are GI fibers.
Using POF made with deuterated polymers would offer a number of advan-
tages. Chemically these materials behave identically to the substances made from
"normal" hydrogen. The attenuation is approximately one order of magnitude less
than the values achieved for PMMA fibers. The behavior over temperature and the
options for index profile design should be the same as those of PMMA-POF.
However, the decisive disadvantage is that there is always water vapor present in
the atmosphere which will be absorbed by the fibers. This will lead to a situation,
where which protons (normal hydrogen nuclei) slowly replace the deuterium so
that the absorption losses will increase again.
Although it is possible to solve the problem with a watertight coating of the
fiber (including all connections), this would defeat the object of obtaining a parti-
cularly low priced cable system.
10,000
attenuation [dB/km]
5,000
2,000
1,000
500 [Koi95]
[Ish92a]
200
[Koi96b]
100 [Koi96d]
50 [Mur96]
20
wavelength [nm]
10
500 600 700 800 900 1000 1100 1200 1300 1400
Fig. 2.177: Loss spectra of GI-POF (deuterated, 1996)
In the past few years work has once again been conducted in Japan on the
production of deuterated POF. GI fibers exclusively have been investigated - see
[Kon02], [Kon03] and [Kon04]. The attenuation of these fibers from [Kon04] is
compared in Fig. 2.178 with the values from 1995 and those of a PMMA-POF.
1500
attenuation [dB/km]
1000 PMMA
500
d8-PMMA 2002
0 1995
450 550 650 750 850
wavelength [nm]
Fig. 2.178: Loss spectra of GI-POF ([Kon02])

Different production versions are compared in [Kon02]. The effect of an

additional PMMA cladding is investigated among other things. The best results
compared with a pure PMMA-POF are shown in Fig. 2.179. With about 60 dB/km
and 650 nm the attenuation ranges approximately between pure PMMA and PF-GI
fibers. On the other hand, the attenuation of the PMMA POF at 520 nm also does
not lie much higher.
10.000
attenuation [dB/km]
5.000
2.000 PMMA
PMMA
1.000
500
200
100 d8-POF
50
20
wavelength [nm]
10
500 550 600 650 700 750 800
Fig. 2.179: Attenuation of deuterated POF ([Kon04])
Since 2003, Fujifilm has been announcing the development of a new fiber
“Lumistar” in the versions I, V and X. According to their own statements this is:
“the first POF with a large diameter which is able to transmit over 1 Gbit/s”. This
is somewhat exaggerated, of course, since PMMA GI-POF and MC-POF have
been able to do this for many years.
power [dB]
0.0
-1.0 1.9 GHz100 m

3 dB-bandwidth
-2.0
-3.0
-4.0
-5.0
frequency [GHz]
-6.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Fig. 2.180: Frequency response of the Lumistar GI-POF ([Nak05b])

Details of a fiber with a core diameter of 500 μm and a cladding diameter of

750 μm are described in [Nak05b]. The bandwidth of the fiber is 1.9 GHz over
100 m. Figure 2.180 shows the frequency response.
Furthermore, the work shows that the index profile produced by gel poly-
merization technology also remains stable after 2000 hours of aging at +90°C
(Fig. 2.181) which is very astonishing.
Fig. 2.181: Refractive index profile of the Lumistar GI-POF after ageing (+90°C)
Parameters for the Lumistar fibers are mentioned in different sources. Accor-
ding to this information a particularly low-attenuation polymer is used. Since the
company works closely with Keio University, where until 2004 there were reports
on the development of deuterated fibers with very similar parameters, we must
assume that we are dealing here with d8 PMMA-POF.
Table 2.24: Data of the d8-POF Lumistar
Lumistar-I Lumistar-V Lumistar-X

core material n. a. n. a. n. a.
core diameter 500 μm 300 μm 120 μm
cladding diameter 750 μm 316 μm 500 μm
attenuation 160 dB/km 180 dB/km <100 dB/km
(650 nm) (650 nm) (850 nm)
bandwidth 1 GHz 50 m 3 GHz 50 m 10 GHz 50 m
According to [Kon05] the fiber is drawn from a 22 mm thick preform (60%

core). By means of a two-stage polymerization process the bandwidth is improved
(Fig. 2.182 shows the losses of two current versions). The optimal NA lies bet-
ween 0.2 and 0.3. The goal is the transmission of at least 3 Gbit/s over 200 m.
1000
attenuation [dB/km]
500
PMMA-d8 core
PMMA-cladding
200
100 PMMA-d8
complete
wavelength [nm]
50
500 550 600 650 700 750 800
Fig. 2.182: d8-POF variants according to [Kon05]

- conventional gel-polymerization with all-PMMA-d8 (79.8 dB/km at 650 nm)
- two level gel-polymerization with PMMA-d8 core and PMMA-cladding (red)
The bandwidth of the fibers was determined through pulse broadening in the
time domain, a SI-POF was used as a mode mixer. The fiber with the PMMA d8
core and the PMMA cladding attains 1.2 GHz · 300 m (overfilled launch). This
version does indeed have a somewhat higher attenuation, but also has a higher
bandwidth due to the index dip at the core-cladding interface.
In October 2004, Fujifilm introduced a DVI transmission system on the basis of
the Lumistar fiber. Using a 850 nm VCSEL a data rate of 10.3 Gbit/s over 40 m
could be transmitted (eye diagram in Fig. 2.183).
Fig. 2.183: 10.3 Gbit/s-data transmission over 40 m PMMA-d8-GI-POF
To what extent this fiber is actually available on the market cannot yet be
assessed since there are no channels of distribution yet in Europe. Even the actual
production costs are still unknown.
The use of fluorine instead of deuterium is indeed more complicated, but does
promise even lower attenuation values and above all long-life fibers. The
following section describes the development of these fibers.
2.7.5 Fluorinated Polymers
The atomic mass of fluorine is many times greater than that of hydrogen so that
the absorption bands are moved significantly further into the infra-red zone. The
theoretical minimum values are less than 0.2 dB/km ([Mur96]), i.e. comparable to
silica fibers in the wavelength range of about 1,500 nm. Figure 2.184 compares
the attenuation values theoretically possible for fluorinated polymers with those
achieved for singlemode glass fibers.
silica glass
PF-polymer
10
0.1 wavelength [nm]

400 600 800 1,000 1,200 1,400 1,600
Fig. 2.184: Theoretical comparison of PF polymer and silica
However, practical experience shows that these impressive theoretical values

are in fact difficult to achieve. The most important question is whether it will be
possible to find a fluorinated polymer that can be processed into a fiber in its
amorphous state. For example, Teflon materials tend to crystallize. Due to scatte-
ring losses, this will significantly reduce the transparency of the material. Even
this very first problem proved to be quite difficult to solve. The second question
relates to the production of the optical waveguide itself. For a step index fiber one
needs a cable material with a slightly smaller refractive index ('n | 0.02 - 0.05).
However, fluorinated polymers already have the lowest refractive index of all
existing transparent plastics (n = 1.340 at 650 nm, or n = 1.336 at 1,300 nm),
which is why they are the preferred material for claddings. The reason why no
PF-SI-POF have been produced to date is simply the fact that there are no suitable
cladding materials available for this purpose.
In principle, graded index POF do not require an optical cladding. On the other
hand, it is necessary to find a way to continually increase the refractive index
towards the axis. Essentially this can be achieved through doping and co-poly-
merization. In the case of silica glass, the index variation can be easily achieved
by replacing the silicon atoms with germanium because these two substances be-
have identically within the glass structure. However, the components used for po-
lymer optical fibers do not allow such a simple replacement of individual atoms.
The process of doping involves inserting small molecules between the long
chains of the actual core material which increases the refractive index. What is
important is that the dopants do not diffuse out of the polymer material too easily
and do not show too strong absorption in the desired wavelength range. The
doping process always lowers the glass transition temperature. It is therefore
desirable to insert a molecule that accomplishes the required change in the refrac-
tive index even at small concentrations (a few percent).
In co-polymerization one uses chains composed of different monomers. The
ratio of monomers determines the refractive index. In this case it is important that
the sequence should be irregular - no long chains of one monomer are formed -
since otherwise the losses due to scattering increase considerably. This means that
the bonding force of monomers amongst each other must not be greater than the
bonding force to the respective other monomer. Of course, both monomers must
have sufficient transparency. Figures 2.185 and 2.186 show a schematic illustra-
tion of the principles.
monomer
dopant
Fig. 2.185: Index variation by dopants
monomer A
monomer B
Fig. 2.186: Index variation by copolymerization

Some fluorine polymers are listed, for example by [Mur96].

¾HFIP 2-FA hexafluoroisopropyl 2-fluoroacrylate
¾PTFE polytetrafluoroethylene
¾FEP tetrafluoroethylene-hexafluoropropylene
¾PFA tetrafluoroethylene-perfluoroalkylvinyl-ether
To date the best results in producing low attenuation POF have been achieved
with the material CYTOP (cyclic transparent optical polymer), developed at
Asahi Glass in Japan. This material no longer contains hydrogen. Its molecular
structure is shown in Figs. 2.187 and 2.188.
CYTOP®
CF2=CF-O-CF2-CF2-CF=CF2
CF2 CF2 CF2 CF2 CF2 CF2

CF CF CF CF CF CF
O CF2 O CF2 O CF2
CF2 CF2 CF2
momomer polymer
Fig. 2.187: Fluoropolymer CYTOP® from Asahi Glass
Fig. 2.188: CYTOP® molecule structure
It was possible to reduce the attenuation of fibers step by step from initially
over 50 dB/km to 30 dB/km and finally to less than 10 dB/km at a wavelength of
1,300 nm, as shown in the data for different PF-GI-POF in Table 2.25.
Different attenuation spectra of GI-POF are compared in Fig. 2.189. The years
indicate the history of the development of this technology. Estimates in [Mur96]
suggest that attenuation for CYTOP will be less than 1 dB/km, bearing in mind
that the need for a GI profile will have a negative effect on this value.
Table 2.25: Data of different PF-GI-POF

Ref. Year Producer Øcore dB/km at nm Remarks
μm
[Koi96c] 1995 Keio Univ. n. a. 50 1300
[Mur96] 1996 Asahi 300- 140 850 n=1.34 (589 nm, Tg=108°C,
Glass Co. 500 56 1300 'n = 0.115, D = 2.4
[Koi96c] 1996 Keio Univ. n. a. n. a. n. a. 10 GHz100 m | 660 nm
[Yos97] 1997 Asahi 125- 56 1300 AN = 0.2, D = 2.4,
Glass Co. 300 nKern = 1.34, 600 MHzkm
[Koi98] 1998 n. a. 40 1300
[Oni98] 1998 Asahi 210 41 850 10,000 h/70°C, AN = 0.18
Glass Co. 45 1300
[Khoe99] 1998 n. a. 120 850
56 1300
[Khoe99] 1998 130 110 650
43.6 840
31 1310
[Koi00] 2000 Asahi 120 15 1300 9 ps/nmkm dispersion
[Kog00] Glass Co. 509 MHzkm@1300 nm
522 MHzkm@850 nm
[Wat03] 2003 Asahi 120 15 1300 till now lowest POF-
Glass Co 8 1070 attenuation
[Gou04] 2003 Nexans 120 40 850 1500 MHz100 m
[Whi04b] 2004 Chromis 120 25 850 400 MHzkm
[DuT07] 2007 Chromis 120 40 800- 800 MHzkm
50 1300 continuously drawn
1,000 attenuation [dB/km]
500
1995
200 1996
100
50
1998
20
2000 wavelength [nm]
10
500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600
Fig. 2.189: Development of the attenuation for PF-GI-POF

Values below 20 dB/km allow transmission ranges of up to 1,000 m. This

covers not only the field of application for copper data cables but also for glass
multimode fibers. Likewise, deployment in access networks would become
possible.
The best values so far are shown in Fig. 2.190 from [Whi02] and [Wat03],
whereby OFS - in the meantime under the company name of Chromis Fiberoptics
- has used a continuous production process for the first time (more information
below).
100
attenuation [dB/km]
60
40
20
10 AGC
OFS
6
600 700 800 900 1,000 1,100 1,200 1,300 1,400
wavelength [nm]
Fig. 2.190: Current attenuation values for PF-GI-POF
2.7.6 Overview over Polymers for POF Jackets
Apart from the materials used in the fiber core, the material used for the jacket is
also important. It has a significant contributing effect on thermal resistance. In
addition, the jacket determines the mechanical properties of the cable such as re-
sistance to compressive load and tensile strength as well as flexibility. Tables 2.26
through 2.30 list different possible materials with some of their characteristics.
The use of PVC, PE or PA as typical jacket materials for applications within
buildings allows for maximum temperatures ranging from 70°C up to 90°C. The
materials in the last two rows (trade names are Teflon FEP or Teflon PTFE) can
be used at significantly higher temperatures.
Table 2.26: Materials for POF jackets (thermal properties)
Short Material VDE- Allowed Conti- Thermal Over-

Name label nuous Operation load Capacity
Temperature 240 h 20 h
PVC polyvinylchloride Y 70°C 80°C 100°C
PVC 90° polyvinylchloride 90°C Y 90°C 100°C 120°C
PVC flame ret. polyvinylchloride flame retardant Y 70°C 80°C 100°C
PE LD; MD polyethylene (low, medium density) 2Y 70°C 100°C 100°C
PE flame ret. polyethylene flame retardant /with halogen 2Y 70°C 100°C 100°C
PE HD polyethylene (high density) 2Y 80°C 110°C 120°C
PP polypropylene 9Y 90°C 110°C 130°C
PA-6 polyamide - 6 4Y 80-90°C 120°C 150°C
PUR polyurethane (thermoplastic) 11Y 90-100°C 120°C 140°C
VPE cross linked polyethylene 2X 90°C 140°C 160°C
EVA ethylene-vinylacetate-copolymere 4G 120°C 160°C 180°C
FEP perfluorethylenpropylene 6Y 180°C 230°C 240°C
PTFE polytetrafluorethylene 5Y 260°C 300°C 310°C
Table 2.27: Materials for POF jackets (thermal/mechanical properties)
Short Proces- Flame Oxygen Thermal Thermal Linear

Name sing *) Resistant Index LOI Value Ho Conductivity Expansion-
MJ·kg-1 W·K-1·m-1 Coefficient K-1
PVC E partly 23-28% O2 17 - 25 0.17 10 - 20·10-5
PVC 90° E partly 23-28% O2 17 - 25 0.17 10 - 20·10-5
PVC flame red. E yes 30-40% O2 15 - 20 0.17 10 - 20·10-5
PE LD; MD E and S no d 22 % O2 42 - 44 0.30 20 - 50·10-5
PE flame red. E and S partly 24-27% O2 35 - 40 0.30 20 - 50·10-5
PE HD E and S no d 22 % O2 42 - 44 0.40 40 - 45·10-5
PP E and S no d 22 % O2 42 - 44 0.19 15·10-5
PA-6 E and S no d 22 % O2 29 - 30 0.23 7 - 10·10-5
PUR E and S no 20-25% O2 23 - 27 0.25 15 - 20·10-5
VPE EoV no d 22 % O2 42 - 44 0.30 20 - 30·10-5
EVA EoV no d 22 % O2 19 - 23 n. a. n. a.
FEP E yes >95 % O2 5 0.26 8 - 11·10-5
PTFE W (E) yes >95 % O2 5 0.26 6 - 15·10-5
*) E: extrusion, S: injection molding, V: vulcanization, W: wrap technology
Table 2.28: Materials for POF jackets (physical/chemical properties)
Short Name Melting Low Tempe- Density Corrosive Harmfull

Agents in the Flue Gas
J-Rays
Temperature rature Limit g·m-3 Resistance
PVC from 130°C -10°C 1.20-1.50 yes d 10 Mrad
PVC 90° from 130°C -10°C 1.20-1.50 yes d 10 Mrad
PVC flame ret. from 130°C -10°C 1.30-1.60 yes d 10 Mrad
PE LD; MD 90-110°C -50°C 0.87 no d 100 Mrad
PE flame ret. from 110°C -50°C 0.98 yes d 50 Mrad

PE HD 125-135°C -50°C 0.95-0.98 no d 100 Mrad
PP from 145°C -20°C 0.91 no d 10 Mrad
PA-6 from 175°C -50°C 1.10-1.15 ? d 10 Mrad
PUR from 150°C -50°C 1.15-1.20 no d 500 Mrad
VPE - -50°C 0.92 no d 100 Mrad
EVA - -50°C 1.30-1.50 no d 100 Mrad
FEP 255-275°C -65°C 2.00-2.30 yes d 0.1 Mrad
PTFE 325-330°C -65°C 2.00-2.30 yes d 0.1 Mrad
Table 2.29: Materials for POF jackets (physical/chemical properties)

Short Oil and fuel Weather Shore-Hardness Tensile Extension
Name resistance Resistance 1) = A; 2) = D Strength Break
PVC middling good 70-951) 10-20 N·mm-2 150-350 %
PVC 90° middling good 70-951) 10-20 N·mm-2 150-350 %
PVC flame ret. middling good 80-901) 10-20 N·mm-2 150-250 %
PE LD; MD bad medium 43-502) 15-20 N·mm-2 300 %
PE flame ret. bad medium 502) 15-20 N·mm-2 300 %
PE HD middling medium 60-622) 15-25 N·mm-2 300 %
PP middling medium 40-602) 30-50 N·mm-2 300 %
PA-6 middling good 40-75 70-120 N·mm-2 50-200 %
PUR good excellent 75-1001) 35-45 N·mm-2 300 %
VPE middl. /good good 40-502) 12-20 N·mm-2 300 %
EVA bad good 70-901) 5-15 N·mm-2 300 %
FEP very good excellent 55-602) 15-25 N·mm-2 250 %
PTFE very good excellent 55-652) 80 N·mm-2 50 %
Table 2.30: Materials for POF jackets (electrical properties)

Short Loss Factor tanG at Permittivity at 20°C Resistivity
Name 20°C and 800 Hz and 800 Hz at 20°C
PVC 20 - 100·10-3 4-6 1013 :·cm
PVC 90° 50 - 100·10-3 4-6 1013 :·cm
PVC flame ret. 70 - 150·10-3 5-7 1013 :·cm
PE LD; MD 0.2; 0.4·10-3 2.3 1016 :·cm
PE flame ret. 1.1·10-3 3 1016 :·cm
PE HD 0.3·10-3 2.3 1016 :·cm
PP 0.5·10-3 2.3 - 2.5 1016 :·cm
PA-6 30 - 50·10-3 3-7 1014 :·cm
PUR 30·10-3 8 1012 :·cm
VPE 0.5·10-3 2.4 - 3.8 1016 :·cm
EVA 20 - 30·10-3 4-6 1012 :·cm
FEP 0.0003·10-3 2.1 1016 :·cm
PTFE 0.0003·10-3 2.1 1017 :·cm
180 2.8 Fiber and Cable Production
2.8 Fiber and Cable Production
The processes for producing POF have been continuously improved in the last few
years. The fundamental methods have indeed always remained the same, but vari-
ous details have been improved. A very comprehensive treatment of POF produc-
tion and its history can be found in [Nal04]. Many fine points concerning the
materials can also be found in [Har99].
As opposed to the production of glass fibers there is a number of unusual
features with POF. First of all, the polymer chemistry involved, in part very com-
plicated and with its occasional safety aspects, has to be mastered. On the other
hand the process temperatures are very much lower - almost always below
+200°C.
The demands on POF production can be sub-divided into four areas:
¾The core material must be produced uniformly without any impurities, air
bubbles, etc. and with a correct distribution of the molecular masses.
¾The fiber must be drawn or extruded exactly.
¾For SI fibers a suitable cladding material with low refractive index and an
attenuation not too high must be found and applied. In doing so, one must
guarantee that the interface is sufficiently smooth and that the cladding has a
good wringing fit.
¾For graded index fibers a copolymer or a dopant must be found in order to be
able to vary - usually increase - the refractive index. A suitable process is
needed in order to distribute this material over the core cross-section so that
you have a parabolic refractive index profile.
There are other steps, of course, such as the application of additional protective
layers, the production of duplex or ribbon cables and quality control.
2.8.1 Production Processes for POF
Today glass fibers are produced in two different ways. The typically 125 μm thin
fibers for telecommunication applications are produced - up to more than 1000 km
- from a preform. Light guiding fibers are drawn directly from molten glass.
Even with polymer fibers one differentiates between continuous methods,
spinning or extruding, and the drawing out of the preform.
In the preform method a cylinder is produced that already has the index profile
of core and cladding while having a much larger diameter. During the drawing
process, the diameter is reduced until the desired size has been reached
(Fig. 2.191, see e.g. [Wei98]).
Ideally, the index profile should be maintained during this process but at a
proportionally reduced scale. The length of the fiber per preform is determined as
follows:
Length of fiber = preform length · (preform diameter/fiber diameter)2
2.8 Fiber and Cable Production 181
This method is applied generally for glass fibers. Automated processes are then
applied to make several 100 km of fiber out of each preform, as the following
example shows:
Length of glass fiber = 2 m preform (5 cm preform diameter/125μm)2 = 320 km
It is easy to see that the large core diameter of common POF is not favorable
for this process since only a few km of fiber can be produced from each preform,
for example:
Length of POF = 1 m preform · (5 cm preform diameter/1 mm)2 = 2.5 km
Drawing speeds for glass fibers today can attain 10 m/s; with POF about 0.2 to
0.5 m/s.
mounting with feed

mechanism
preform
oven
take up drum
diameter control
unit
Fig. 2.191: Production of POF from a preform
In addition to being able to draw the complete fiber out of the preform there is
also the possibility of producing the core as a polymer cylinder and then applying
the cladding by extrusion or enameling. The advantage here is that the polymeri-
zation of the core material can proceed under very much better controlled
conditions.
This process is used with PCS. A silica glass core is drawn out to 200 μm - or
to other thicknesses as well - and is then surrounded by a polymer cladding,
typically 15 μm thick. Understandably, the glass and the polymer have to be pro-
cessed using different procedures.
Other versions are discontinuous production in which polymerization first takes
place in the reactor and then the resulting block is extruded at low temperature, a
so-called batch extrusion.
N2
vacuum-
pump
reactor
mixer
heater
monomer initiator cooler
polymerization
controller
cladding
polymer
POF with cladding
Fig. 2.192: Batch-extrusion according to [Hess04]
The monomer, the initior and the polymerization controller are first distilled by
a vacuum pump. After the polymerization is finished, nitrogen pushes the poly-
mers through the nozzle and the cladding is then immediately applied.
In addition, Mitsubishi has developed a method with which the polymerization,
described in [Nal04], can take place photochemically.
Figure 2.193 from [Hess04] shows such a method. The core and cladding
materials are pushed through a nozzle by a pump and a mixer. The cross-linking
then takes place with a UV lamp. This process could prove to be quite suitable,
especially for heat-resistant POF.
mixture
spinning
nozzle
cladding core take up
material material drum
UV-light for
crosslinking
Fig. 2.193: Polymer crosslinking

When extrusion techniques are applied, the POF is produced in a continuous

process directly from monomers. For SI-POF this process is very simple.
Figure 2.194 shows such an arrangement (e.g. [Ram99], [Wei98]).
POLYMER
filler heated
vessel
diameter
core control
extruder fiber
conveyor
pump
cladding
extruder
Fig. 2.194: Production of SI-POF through extrusion
Such a system is also described in [Hac01]. According to the author the

cladding materials used are Poly(3FMA) with n = 1.40 and PVF with n = 1.42.
The polymerization takes place at about 150°C. With the drop in pressure when
leaving the reactor the remaining monomer is vaporized and can be returned. The
cladding is extruded at about +200°C. This temperature lies far above the glass
transition temperature for PMMA. Thus is a critical step in the process in which
the quick cooling of the fiber must be guaranteed. On the other hand, the cladding
is only about 10 μm thick so that the thermal load is limited.
monomer, initiator,
polymerization controller
reactor
cladding extruder
heating
fiber
pump
extruder
take up drum
Fig. 2.195: Extrusion of a POF according to [Hac01]

This process is also discussed in [Hess04]. The monomer is polymerized to

about 80% in the reactor. The advantage of this standard process for SI-POF lies
in the very slight contamination of the polymers caused by the process.
A modification of the process is presented in [Poi06d]. The new components in
the process are:
¾The core material is PMMA granulate which is crushed before extrusion and
effectively cleaned.
¾The extrusion head is to be kept free of metal and any impurities whatsoever
if possible.
¾The turbo pump used makes a particularly even transport possible.
In addition, two further processes are mentioned in [Wei98]. In the thrust extru-
sion technique, polymerization is carried out in a closed heated container from
which the fiber is subsequently expelled through a nozzle at high pressure. The
cladding is applied directly within the nozzle. This is a non-continuous process
just like the preform technique.
In the spin-melt process, a volume of ready-to-use polymer pellets is melted
and pressed through a spin head that incorporates many holes. The holes serve to
form the core and apply the cladding. This process is very efficient but also very
expensive.
2.8.2 Production of Graded Index Profiles
In order to guarantee the optimal functioning of graded index and multi-step index
fibers, the best index profile possible should be realized. The developmental goal
of the past few years has been to attain as much as possible with minimum effort
and to continuously produce GI fibers.
A number of different processes for the manufacture of graded profiles are
described in the technical literature:
¾Interfacial gel polymerization technique
¾Centrifuging
¾Photo-chemical reactions
¾Extrusion of many layers
In most of these techniques the principle is to initially create a preform of up to
50 mm diameter and then to subsequently draw this preform down to the desired
fiber size. Some of these methods are described below.
2.8.2.1 Interfacial Gel Polymerization Technique

This method was developed by Prof. Koike of the Keio University (for an example
see [Koi92]). In this process a tube is initially manufactured with PMMA. This
tube is then filled with a mixture of two different monomers M1 (high refractive
index and large molecules) and M2 (smaller refractive index and smaller mole-
cules). Initially the inner wall of the PMMA tube is slightly liquefied in an oven
that has been typically heated to 80°C. This results in a layer of gel and accele-
rates polymerization. The smaller molecule M1 can more easily diffuse into this
layer of gel so that the concentration of M2 increases more and more towards the
middle. The index profile is thus formed in accordance with the resulting concen-
tration gradient. For manufacturing a PMMA-GI-POF, [Koi92] proposes that
MMA (M1) be supplemented with monomers VB, VPAc, BzA, PhMA and BzMA.
The material that was finally used is BzA because its reactivity is comparable with
that of MMA. The 15 mm - 22 mm thick preform is then drawn at temperatures
between 190°C and 280°C to produce fibers ranging from of 0.2 mm - 1.5 mm in
diameter. Figure 2.196 illustrates the principle (see also [Ish95]).
PMMA tube 80°C

filled with a melting of the PMMA tube
MMA/BzA mix formation of a gel layer
the gel layer moves to the center

concentration of M2 increases from outer to the the center
Fig. 2.196: GI profile formation by gel polymerization technique
[Koi95] describes this method in more detail. The PMMA tube is produced by
rotating a glass reactor at 3,000 min-1 at 70°C that is partially filled with MMA.
The polymerization process for the core takes place at a speed of 50 min-1 and a
temperature of 95°C and requires approximately 24 hours to complete. [Ish95]
describes the production of a PMMA GI-POF with DPS as dopants. For traditional
materials such as BB or BBP, one obtains fibers with a NA of 0.17 - 0.21, whereas
with DPS a NA of 0.29 is possible. The greater NA improves the bending charac-
teristics and makes the launching of light easier.
2.8.2.2 Creating the Index Profiles by Centrifuging

Several publications ([Dui96], [Dui98] and [Chen00]) propose utilizing the den-
sity difference of the different monomers to create the index profile through cen-
trifugal force in a fast centrifugal process. [Chen00] compares the density and
refractive index of different materials for this purpose (Table 2.31).
Table 2.31: Refractive index and density of different polymers ([Chen00])
Molecule Density n Molecule Density n

-3 -3
MMA 0.936 g/cm 1.490 BB 1.120 g/cm 1.568
DOP 0.981 g/cm-3 1.486 PMMA 1.190 g/cm-3 1.490
-3 -3
BIE 0.982 g/cm 1.564 TFPMA 1.254 g/cm 1.373
-3 -3
BzMA 1.040 g/cm 1.568 PTFPMA 1.496 g/cm 1.422
VB 1.070 g/cm-3 1.578 DBME 2.180 g/cm-3 1.538
The production of the preform is carried out in two steps. Once the monomer
mixture has been filled into a tube, the GI profile is formed at room temperature.
Then the temperature is increased so that polymerization takes place. Rotation
continues during this process. Then the fiber is drawn from this preform.
In this process the rotation speeds must be up to 50,000 min-1. Even for a pre-
form with 10 mm diameter the centrifugal acceleration (a = M2r) already equals
14,000 times the acceleration due to gravity. At the University of Eindhoven an
ultra centrifuge operating at 50,000 min-1 has been constructed for preforms up to
50 mm in diameter which produce a centrifugal acceleration of 70,000 g. In the
first trials, GI cylinders were produced from PTFPMA and MMA. The process for
forming the GI profile took 24 hours. This was followed by a period of 12 hours
during which the polymerization process was carried out at 60°C to 80°C. The
refractive index difference achieved was approximately 0.009. No research reports
have as yet been published on the production of fibers from such preforms.
2.8.2.3 Combined Diffusion and Rotation
The combination of diffusion and rotation for producing PMMA-GI preforms is

described in [Park01]. The monomer is filled into a cylindrical glass reactor in the
middle of which a rod made of a material with a high refractive index is located.
This material diffuses slowly into the surrounding medium. Both parts can rotate
at different speeds: the reactor at 500 to 1000 RPM and the rod at 6 to 60 RPM.
The idea for different rotation speeds comes from determining the average of
concentration fluctuations so that an ideal rotation-symmetrical profile comes
about. After a few hours the preform is thermally polymerized. Figure 2.197
shows the principle and an index profile.
A fiber with a 1 mm core diameter was produced through thermal drawing from
the preform described above. The bandwidth-length product amounts to 1.2
GHz · 100 m, measured with a 650 nm InGaAsP laser on a 50 m long fiber.
solid copolymer with concentration

higher refractive index 1.0
phase 1 phase 2 after 5
room heated hours
tempe- 0.8
rature
0.6
poly-
laminar
meri-
mix of the
zation 0.4
phases
rotating reactor with copolymer is final
monomer, initiator diffused into the GI-preform
and polymerization monomer mix 0.2
controller
n n n 0.0
0.0 0.2 0.4 0.6 0.8 1.0
relative radius
Fig. 2.197: Fabrication of GI-POF-preforms according [Park01]
2.8.2.4 Photochemical Generation of the Index Profile

According to [Nal04] the first GI-POFs were also produced by photo-copoly-
merization, introduced in 1981 by Koike. A thin glass tube is filled with a mixture
of MMA, vinylbenzoate (VB as dopant) and benzoyl peroxide (as initiator). The
glass tube rotates during the UV irradiation. Since the UV radiation is higher at
the edge a gel phase forms here through faster polymerization. The VB concentra-
tion will be greater in the center since MMA has a faster reaction speed. The tube
is irradiated from bottom to top and then polymerized out at high temperatures.
This procedure did not result in any usable fibers.
[Miy99] proposes a method for the production of index profiles by means of a
photo-chemical reaction. In this process, PMMA is doped with DMAPN ((4-N,N-
dimethylaminophenyl)-N’-phenylnitrone). During exposure to ultra violet radia-
tion (380 nm) the refractive index is reduced by up to 0.028, sufficient for
GI-POF. In the experiment, thin films of a few micrometer thickness were used.
Fibers have not yet been produced. It is likely that a problem would be the depth
of penetration of the radiation which is significantly less than the intended fiber
radius. Nevertheless, this process is of great interest since it works fast and makes
continuous fiber production possible.
2.8.2.5 Extrusion of Many Layers

This multi-step index POF has hitherto been produced at two institutes (Research-
Production Center, RPC Tver) and Mitsubishi Rayon.
The process corresponds to the production of SI-POF or DSI-POF except that
several extruders must be combined with one another. Figure 2.198 shows the
index profile of an MSI-POF according to [Lev99]. The curve drawn corresponds
to that of an ideal parabola. In the core area the deviations of the real structure are
relatively small.
refractive index [a.u.]

60
50
40
30
20
10
0
0 100 200 300 400 500
distance to the fiber axis [μm]
Fig. 2.198: Index profile of a MSI-POF ([Lev99])
2.8.2.6 Production of Semi-GI-PCS

The production of the preform for semi-GI-PCS does not in effect differ from the
manufacturing methods for normal glass fibers. The usual process is MCVD (mo-
dified chemical vapor deposition). A mixture of SiCl4 and O2 are introduced into a
heated quartz glass tube and SiO2 is formed by the chemical reaction. By adding
chlorine, boron, germanium or phosphorus, you can continuously change the re-
fractive index (Fig. 2.199). After cooling off, the tube with the inner layer will be
collapsed, i.e. the hole disappears, and is drawn into a fiber. As opposed to classic
glass fibers the PCS has an optical cladding made of polymers, not of glass, thus
making a considerably greater refractive index jump possible.
O2
porous preform
burner
ceramic or
SiCl4 rail graphite rod
gas mixing
controller
GeCl4
Fig. 2.199: Fabrication of glass fiber preforms (by OVD)

2.8.2.7 Polymerization in a Centrifuge

A new method for producing PMMA GI-POF ready for production has been deve-
loped over the past few years by the South Korean company Optimedia under the
direction of Prof. C. W. Park.
The production principle is based on copolymerization. As opposed to doping
there is the advantage of the glass transition temperature not dropping as much.
The polymer mixture is filled into a rotating tube and polymerized thermally or by
UV irradiation. The polymer composition can be changed in steps or continuously.
The rotation here does not serve the purpose of separating the materials, but only
for achieving rotational symmetry. There are correspondingly fewer demands on
the rotation speed. Figure 2.200 shows the set-up. A detailed description can be
found in [Park06a].
Fig. 2.200: Rotating cylinder for GI-preform fabrication ([Park06a])
You can see quite well under a microscope that the fiber is built up of many
layers. Nevertheless, the index profile is almost ideally parabolic and does not
show any steps - see Fig. 2.201 acc. to [Park06a]. An attenuation spectrum of the
OM-Giga, 1 mm GI-POF (data provided by the distributor Fiberfin) is shown in
Fig. 2.202. At 650 nm the losses are below 200 dB/km.
1.525 refractive index

1.520
1.515
1.510
1.505
1.500
AN: 0.30
1.495
1.490 normalized radius [mm]
0.0 0.1 0.2 0.3 0.4 0.5
Fig. 2.201: Refractive index profile of a PMMA-GI-POF made by Optimedia ([Park06a])

2000
1000
500
200
wavelength [nm]
100
400 500 600 700 800 900
Fig. 2.202: Attenuation spectrum of a PMMA GI-POF made by Optimedia (Fiberfin)
2.8.2.8 Continuous Production at Chromis Fiberoptics

While there are continuous production processes for SI-POF, PF-GI-POF could
only be produced until just recently from preforms. Chromis Fiberoptics - pre-
viously Lucent, OFS - has developed a process for the continuous production of
such fibers ([Rat03], [Whi03], [Whi04a], [Whi05], [Park05b] and [Pol06a]). First
a SI fiber of CYTOP material with a doped core is produced in a double extruder.
The fiber is wound around a heated cylinder. Here the dopant diffuses outwardly
resulting in the GI profile. The 500 μm PMMA protective layer is then applied
and the fiber can be wound up. The fibers almost attain the parameters of POF
from Asahi Glass which has had about 10 years of experience in the field.
cladding core extruder step-

extruder (CYTOP + dopant) index
(CYTOP) profile
dopant
diffusion
coextrusion
head GI-POF indexprofil
protective heated tube index
layer difference
extruder
coextru-
sion head
diameter
capstan
control
-100 -50 0 50 100
radius (μm)
to the take up drum
Fig. 2.203: Continuous PF-GI-POF fabrication ([Pol06a])

The insert shows the final index profile with an approximately parabolic curve.
The manufacturer indicates the bandwidth-length product of the fiber as being
400 MHz · km.
2.8.2.9 GI-POF with Additional Cladding

As already indicated above, a reduction in the bending losses plays a great role
with polymer fibers. For SI fibers a considerable improvement could be achieved
by means of a second cladding. Extremely small bending radii can be attained
through fiber bundles or multi-core fibers respectively.
For graded index fibers as well, an additional cladding layer with a smaller
refractive index evidently offers clear advantages in regard to the bending beha-
vior without dramatically reducing the bandwidth. A PF-GI-POF with an addi-
tional 6 μm thick cladding layer is introduced in [Oni04] and [Sato05].
Figure 2.204 shows the measured bending losses for three different fibers with
different index jumps between the edge of the core and the additional cladding
(around ¨n = 0.002, ¨n = 0.005 and ¨n = 0.014). Even with an index jump of
0.005 a bending radius of 10 mm with an attenuation below 0.1 dB can be
attained. The bandwidth-length product of the fiber lies between 1,800 MHz · km
and 2,700 MHz · km. The fiber attenuation amounts to 30 dB/km at 850 nm,
measured with ODTR.
180°-bending loss [dB]

1.2
1.355
n
1.350
1.0
1.345
0.8 1.340
1.335
0.6 1.330
x [μm]
1.325
0.4 0 20 40 60 80 100 120 140 160 180 200
0.2
bending radius [mm]
0.0
0 10 20 30 40 50 60
Fig. 2.204: Reduction of the bending losses due to a Semi-GI profile ([Sato05])
This method can also be employed for PMMA-GI fibers. The results for a
1 mm thick fiber are presented in [Aru05]. The attainable bending radius drops to
below 5 mm with an additional PVDF cladding (polyvinylidene fluoride,
n = 1.42). The bandwidth-length product of the fiber is 1,500 MHz · 100 m and re-
mains quite constant up to 10 mm. It only drops under full launch and with a
5 mm bending radius to 500 MHz · 100 m. The attenuation at a 90° bend is com-
pared to a conventional PMMA GI-POF in Fig. 2.205.
2.0
bend loss [dB]
PVDF clad GI-POF
NA of the GI core region = 0.17
1.5
PMMA based GI-POF
NA of the GI core = 0.21
1.0
0.5
0.0
bend radius [mm]
-0.5
0 10 20 30 40 50 f
Fig. 2.205: Bend losses in Semi-GI-POF according to [Aru05]
In addition to the extra cladding layer a so-called W-profile for GI fibers has
also been developed. Here the goal is to improve the attainable bandwidth.
Measurements on PMMA GI-POF with this W-profile and different index expo-
nents are presented in [Tak05b]. The W-profile is characterized by a very steep
index drop directly at the core-cladding interface. Figure 2.206 shows the index
curve.
Fig. 2.206: W-profile for PMMA-GI fibers ([Tak05b])
Furthermore, fibers with a NA of 0.20 and a ȡ-parameter (index exponent of the

rise outside the core-cladding interface layer) have been produced with index
exponents between 1.9 and 5.2. Figure 2.207 shows the theoretically calculated
and measured bandwidths.
5.0
3 dB bandwidth [GHz100 m]
3.0 W-shaped POF
2.0
1.0
0.5 GI-POF
0.3 calculated for GI-POF
0.2
1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5
profile index exponent g
Fig. 2.207: Bandwidths of PMMA-GI-POF, improvement by W-profile [Tak05b]
PF-GI-POF with optimized index profiles are presented in [Ebi05]. Their band-
width attain that of MM-GOF and in the short-wave range even surpasses it (Table
2.32). The high bandwidth is attained through the approximately ideal index
coefficients of 2.05, i.e. in combination with the low chromatic dispersion of the
material.
Table 2.32: Bandwidths comparison of GI-GOF and POF according to [Ebi05]
Bandwidth
wavelength 650 nm 780 nm 850 nm
PF GI-POF 8.39 GHz 8.50 GHz 9.54 GHz
SiO2-GI-GOF 5.27 GHz 7.34 GHz 9.31 GHz
Figure 2.208 shows the best attenuation values over time for some of the fibers
listed above. PMMA fibers (SI and GI) reached their theoretically maximum
possibilities in the mid-80s. Since then, other index profiles (MSI, MC, DSI) have
also reached this order of magnitude (approx. 130 dB/km at 650 nm and 80 dB/km
at 570 nm). Any differences in measured values and specifications are more likely
to result from different measuring conditions than from differences in quality.
The PF fibers have been continually improved, at least as far as the laboratory
results are concerned. The best values were attained in 2003 with about 8 dB,
almost one magnitude still above the theoretical limits. In the past three years no
further progress has been made with the attenuation. On the other hand, there has
been some success in attaining a high launch-independent bandwidth with opti-
mized refractive index profiles and in reducing the bending sensitivity.
attenuation [dB/km]
1,000
500
GI-PMMA
200
100
50
SI-PMMA at 650 nm
20 SI-PMMA at 570 nm
SI-d8 at 680 nm
PF-GI at 1.300 nm
10
d8-GI at 688 nm
5
1960 1965 1970 1975 1980 1985 1990 1995 2000 2005
year
Fig. 2.208: Development of POF attenuation until the year 2005
2.8.3 Cable Manufacturing
This chapter discusses the structure and properties of various cable structures with
POF wires. Different applications place different demands on the mechanical
shielding of the polymer optical fiber. SI-POF (Step Index Polymer Optical Fiber)
is a promising medium for relatively short transmission distances of 100 m. Poly-
mer plastics such as polymethylmethacrylate (PMMA) or polycarbonate (PC) are
used as the primary core material for manufacturing these fibers. Fluorinated poly-
mers, silicone or fluorinated PMMA materials are used as cladding material with a
reduced refractive index of ncladding ~ 1.42 as compared with the core material
ncore > 1.48 (Fig. 2.209).
Due to the large refractive index difference, numerical apertures of up to 0.50
are attained. Various manufacturer versions of optical fibers are shown in
Fig. 2.210, in which glass or plastic are combined for the core and cladding mate-
rial. The relatively thin glass fibers are mechanically fragile and must therefore be
protected by a multilayer cable construction. The POF is so flexible that a simple
jacketing of the optical cladding suffices as a cable construction.
d 0.98 mm r ncladding
D 1.00 mm ncore D
ncore 1.492 d
ncladding 1.416 n
NA 0.47
core material: Polymethylmethacrylat (PMMA)

cladding material: fluorinated PMMA
Fig. 2.209: Typical SI-POF parameters
Glass fibers with polymer optical cladding represent an intermediate step. They
also have a relatively simple construction (two-layer plastic coating around the
optical cladding). The large core diameter allows only step-index profiles.
singlemode glass fiber

optical core
optical cladding
primary coating
10/ 125/ 250 μm
multimode glass fiber secondary coating
strength member
outer jacket
glass fiber with polymer cladding50/ 125/ 250 μm
polymer fiber 200/ 230 μm
980/ 1000 μm
0 mm 0.5 mm 1.0 mm
Fig. 2.210: Comparison of different kinds of optical fibers

Until recently, step-index profile fibers were manufactured almost exclusively

from polymer plastics having a typical outer diameter of 1 mm. These SI-POF
exhibit significant transmission ranges with a minimum of attenuation for wave-
lengths between 400 nm and 900 nm (Fig. 2.211).
The effective spectral loss windows are at 520 nm, 570 nm, 650 nm, and
760 nm. With improved purity, homogeneity and deuterated or fluorinated poly-
mers, it is possible to reduce attenuation to 10 dB/km, as has already been des-
cribed in Chapter 2.7.5.
10,000
attenuation [dB/km] attenuation
5,000 minimum
PC
2,000
1,000
500
PMMA
200
100
50
450 500 550 600 650 700 750 800 850
wavelength [nm]
Fig. 2.211: Attenuation spectrum of different POF made from PMMA or PC
Polymer optical fibers that are flexible and break-resistant can be produced
with a relatively large diameter (up to 1.5 mm or even more) and are thus easy to
handle and to install. The large core diameters in combination with the numerical
aperture make simple connection fittings and equipment possible with low de-
mands on precision.
2.8.3.1 Cable Construction with SI-POF Elements

SI-POF cables or lines must always be flexible when laid/installed at the end user
place. SI-POF must also be flexible for mobile applications.
The flexibility of a line or cable depends on the number and dimensions of the
stranding units with the number of the layer changes of the individual stranding
elements. The shorter the pitch length is and the larger the number of layer chan-
ges, the larger the flexibility of the stranding unit. The pitch length of the indivi-
dual POF wires or the stranding elements with the proper diameter has a major
influence on the flexibility of the stranding elements. The shorter the pitch length,
the more flexible the stranding unit will be (Fig. 2.212).
flexibility
length of lay
Fig. 2.212: Schematic diagram of the relationship between the pitch length and the flexi-
bility of the stranding construction
2.8.3.2 Non-Stranded SI-POF Cables

SI-POF Simplex Cable
When processed into a cable with the respective strain relief, the SI-POF can be
coated with a diffusion lock made from metal over the first cladding, if required.
An absolute diffusion lock can be attained exclusively with a closed tube, for
example with laser-welded metal tubing. The metal strip material for laser welding
can be made of aluminum, copper or high-grade steel. The foil thickness is typi-
cally between 50 μm and 150 μm for welding. For overlapping with or without
gluing, the metal foils have a sandwich layer construction, i.e., 9 μm / 20 μm /
9 μm = metal / plastic carrier strip/ metal.
A jacked is extruded onto the traction elements in combination with the metal
diffusion locks. This layer is practically always flexible and sturdy; polyurethane
or polyethylene are the preferred materials. The next illustration (Fig. 2.213)
shows two typical SI-POF simplex cable constructions.
outer sheath
metal band
inner coating
cladding
fiber core
2.2 mm 2.3 mm
Fig. 2.213: Structure of optical fibers with internal cladding

SI-POF Duplex Cable

The simplest form of a duplex cable is the combination of two parallel POF wires
that are protected by shielding and equipped with traction elements. Various con-
struction options for a duplex cable or duplex line are possible. Two very well
known cable constructions are shown in Fig. 2.214.
foil tape lapping

POF-element
inner coating
outer sheath
strain relief element/
2.5 mm
5 mm rip cord 5 mm
Fig. 2.214: SI-POF duplex cable in a round cable and flat cable form
With these duplex cable constructions, care must be taken to ensure that the
strain-relief elements in the plugs or on the connectors are included in processing.
This is necessary because the temperature influence on the SI-POF wires is con-
structed in such a way that optimum temperature characteristics are ensured in the
temperature range from -40°C through +80°C.
SI-POF and GI-POF Ribbon Cable

A ribbon cable with n SI-POF elements can be constructed as an extension to a
duplex cable. The SI-POF elements are lined up in parallel as a comb and com-
bined in either groups of 5 or 10 elements. A thin protective coating is extruded
over this ribbon cable with the respective traction and support elements in one
work cycle. Various SI-POF ribbon cable constructions with a modular design are
illustrated in Fig. 2.215.
twin group 5 cables group

outer sheath
strain relief element
POF
2.3 2.3 13 mm
inner coating
5.2
10 cables group
26 mm
Fig. 2.215: SI-POF ribbon cable with traction and support elements
The cross-sections of two POF ribbon cables from [Boc04] are shown in
Fig. 2.216. The individual fibers have each been extruded in a joint acrylic
cladding.
Fig. 2.216: Ribbon with four 500 μm SI-POF (above) und eight 120 μm/500 μm GI-POF
(below, [Boc04])
For the OVAL project (see Chap. 6) of the POF-AC Nürnberg Nexans had pro-
duced 8-strand ribbon cables made of SI- and GI-POF each with a 500 μm dia-
meter. The cross-section of a prototype with PMMA-GI-POF (Optimedia) is
shown in Fig. 2.217.
Fig. 2.217: POF-ribbon cable with eight 500 μm OM-Giga-fibers (dimensioning in μm)
The spacing between the individual fibers deviates only slightly from 500 μm.
Only in a vertical position great deviations do arise which can easily be avoided
by better guiding of the individual fibers in the extrusion tool.
In order to investigate the influence of the ribbon cable production on the
optical parameters, the spectral attenuation and the bandwidth were determined on
the SI-POF ribbon cables. The results are shown in Fig. 2.218 and 2.219.
1000
attenuation [dB/km]
800
600
400 fiber 1 fiber 2

fiber 3 fiber 4
300 fiber 5 fiber 6
fiber 7 fiber 8
200
wavelength [nm]
100
450 500 550 600 650 700 750 800
Fig. 2.218: Single fiber attenuation in the ribbon cable
The attenuations of the 8 fibers agreed within the usual measurement error of
±0.5 dB. There were also no significant deviations in the frequency response in
Fig. 2.219.
+5
rel. level [dB]
0
-5
-10
-15
-20
-25
-30
frequency [MHz]
-35
1 3 10 30 100 300 1000
Fig. 2.219: Frequency response of the fibers in the ribbon cable
In one last experiment we investigated whether the ribbon cable production

increased the mode mixture in the fibers. We determined the far field width of the
individual fibers and ribbon cables with under filled launch for different lengths.
The experiment on the ribbon cables was repeated after annealing (120 min. at
+90°C) and aging (200 hours). As can be seen in Fig. 2.220, it took practically the
same length of time in all four cases to achieve equilibrium mode distribution. In
other words, the ribbon cables did not influence the mode mixing processes.
36
FWHMeff [°] high NA-fiber
NAlaunch = 0.10
32
28
24 fiber
ribbon cable
20 annealed
aged
16
lPOF [m]
12
0.1 0.3 1.0 3.0 10.0 30.0 100.0
Fig. 2.220: Effect of mode mixing in fiber ribbons ([Har06])
SI-POF Hybrid Cable

Hybrid cables are characterized by the fact that they are constructed from a combi-
nation of SI-POF elements with cooper-insulated wires that can be joined together
individually or in pairs.
Furthermore, there are hybrid cable combinations in a coaxial construction with
a metallic tube, the so-called POF-CMT element (CMT = Corrugated Micro
Tube). The illustrations shown in Fig. 2.221 point out possible combinations with
SI-POF copper elements or SI-POF aluminum elements in a coaxial construction.
basic element hybrid-cable
POF CMT single duplex triax
isolation
2 ... 3 mm 3 ... 4 mm 7 ... 9 mm 4 ... 5 mm
Fig. 2.221: New design for POF with CMT as electrical conductor
The advantage of such hybrid cable constructions is the possibility of supplying

current directly to the transmitter and/or receiver of the individual SI-POF ele-
ments ([Ziem99a], [Ziem99b]). The connector combination for hybrid cable con-
structions are well known and are used in the automotive field.
Apart from the coaxial hybrid solution, the layer-stranded hybrid solution is
also well known (Fig. 2.222 and 2.223).
copper wire
POF
980/1000 μm
foil
support element
POF
980/ strain element
1000 μm
copper wire
inner coating
inner coating
outer sheath
6.5 mm outer sheath
7.5 mm
Fig. 2.222: Layer-stranded POF-Cu cables (principle)
In these cases, insulated copper wires and POF wires are processed either into a
group of four or as stranded layers with several stranding elements. The copper
wires are used with diameters of 0.5 mm to 1.5 mm. Thicker copper wires are pro-
cessed as braided wires, because the flexibility of the cable usually does not meet
customers’ requirements.
Fig. 2.223: Hybrid POF-Copper cable
2.8.3.3 Stranded SI-POF Cables

Introduction
SI-POF cables or SI-POF lines are products that must bend easily when they are
used and when they are processed. This requirement must be met for the manu-
facturing process or for transport purposes or for winding the cables or lines on
production-machine reels or shipping reels or when sold in rings. The individual
SI-POF elements are twisted in a screw-like fashion around an imaginary center-
line. Twisting is necessary in order for the manufactured products to be flexible
and portable.
The advantage of twisting is that the stranding element is stretched and com-
pressed alternatively on the inner and outer side of a curved section (Fig. 2.224). If
the section in which a SI-POF stranding element is wrapped 360° around a twist
axis that is considerably smaller than the curved section, the strain and pressure in
a stranded construction are constant and it is possible to bend this SI-POF cable
without deformation.
Fig. 2.224: Comparison of cable constructions with short or long lay lengths in terms of the
bending characteristics
The flexibility of an SI-POF cable or SI-POF line is a function of the geometric

dimension of the stranding elements and of the change in layers present in a cable
construction. For example, a large number of layer changes results in a greater
flexibility of the SI-POF cable construction.
The SI-POF stranding elements are wrapped spirally around the twist axis in
various machine configurations. The foundation for these various machine designs
is always the result of a rotary motion with a linear motion. This can be seen
schematically in Fig. 2.225.
1
2 3 5
s 4
n1 d n2
DA
1. rotor s: pitch length d: diameter of the
2. stranding elements n1: rotational speed of stranding unit
3. stranding unit the stranding basket n2: rotational direction
4. capstan gear DA : diameter of the and speed of the
5. stranding axis stranding basket capstan gear
Fig. 2.225: Schematic diagram of the spiral-shaped strands
The option of being able to twist SI-POF elements together is determined by

the following parameters.
¾Pitch length
¾Lay direction
¾Multiplication factor
¾Number of strands
2.8.3.4 Principles of Stranding

Pitch Length
The pitch length is the distance between two points on the twist axis. Within these
two points, the SI-POF element has been rotated 360° around the twist axis. The
lay length is calculated from the following variables:
D A S n2 v m 1000
s [mm] s
n1 n1
where DA: Diameter of the capstan gear

n1: Rotational speed of the stranded basket
n2: Rotational speed of the capstan gear
vm: The machine’s pull-off speed
During the manufacture of twisted SI-POF cables or SI-POF lines, the lay
length s must be determined very exactly because of the precise geometry in-
volved. This means that for stranding machines for SI-POF elements that are
twisted via a capstan gear or caterpillar, the diameter of the stranding elements
must be taken into account. In practice, a deviating diameter for the SI-POF stran-
ding construction is the result and increases the pitch length manufactured. The
geometric assignment is easy to see in the enclosed illustration (Fig. 2.226); the
manufacturing pitch length SH is calculated from it.
d
3
1 1: capstan gear
2: fiber loop
3: POF
2
DA
Fig. 2.226: Diagram for explaining the concept of 'manufacturing pitch length'
The manufacturing pitch length is calculated from the following parameters:

D d
sH s A
DA
sH: Manufactured pitch length

s: Pitch length in machines
DA: Diameter of the capstan gear
d: Diameter of the stranded unit
Lay Direction
The rotational direction of the stranding basket determines the lay direction. The
following distinction is made depending on the sense of direction of the helix:
¾Z-lay means a right-handed thread
¾S-lay means a left-handed thread (Fig. 2.227)
Fig. 2.227: Schematic explanation of the lay direction

The following diagram (Fig. 2.228) illustrates how an SZ stranding is to be

interpreted. It can be seen that, after a number of rotations, the lay direction is
changed. In contrast to classic basket stranding, SZ stranding has the advantage of
having a pull-off speed that is 5-20 times faster.
S Z S Z S
Fig. 2.228: Explanation of the lay direction schematically
Economic and engineering stranded cable products are manufactured exclu-

sively using the SZ stranding method, i.e. also for POF applications.
In classic production, SI-POF stranding elements constructed from several
stranding layers are given alternatively a Z and an S direction. This cable con-
struction element - the SZ-stranding method - for SI-POF results in a very com-
pact geometric shape of the stranding construction, which allows it to cushion well
both traverse and longitudinal forces. The stranding element is to ensure that the
optical transmission values are retained during the manufacturing process of the
cable product and to ensure that there are no changes after laying the cables and in
subsequent operation.
Multiplication Factor
The helical SI-POF stranding element (Fig. 2.228) is longer in the stranded unit.
The stranding method always leads to an increase in material consumption. The
ratio of the laid length L of the SI-POF stranding element to the lay length of the
stranded unit results in the well-known multiplication factor f = L/s. The multi-
plication factor f is determined from the pitch length and the average diameter Dm
in the stranding layer.
The multiplication factor can be easily derived from the triangle shown in
Fig. 2.229.
2
L ( S Dm )2 s2 § S Dm ·
L ( S Dm )2 s2 and f ¨ ¸ 1
s s © s ¹
with L: Laid length L = s/cos Z
f: Multiplication factor
Dm: Average diameter of the stranded layer
s: Pitch length of each stranded layer
For relatively large pitch lengths (Dm « s), the calculation can be simplified as
follows:
f | 1 S Dm / s 2 /2
L SDm
s
Dm Z
s D
Fig. 2.229: Graphical representation of the SI-POF stranding element
Number of Strands
To characterize the bending properties of an SI-POF stranding element v, the
number of strands is formed from the quotient of the pitch length and the average
diameter Dm (v = s/Dm).
s: Pitch length of each stranded layer
Dm: average diameter of this stranded layer
v: Number of strands
Production developments in stranded cable constructions or SI-POF cable con-
structions have lead to the number of strands being v > 8. By using the number of
strands v, the multiplication factor f can be easily calculated.
2
§ S· S2 v 2
f ¨ ¸ 1 | 1 S²/2v²
©v¹ v
Layer Structure
Standard SI-POF elements have a simple geometric shape but have an exact dia-
meter. This makes it easy to calculate SI-POF cables or SI-POF lines. An SI-POF
cable in its classic form, i.e. with a core element, has the same diameter as the SI-
POF element; it can be constructed in a circular fashion with 6 SI-POF elements in
the same layer. The cladding lines are in contact with each other. Two different
core layers have been adopted schematically in Fig. 2.230. The other layers are
calculated and shown. In Table 2.33 and Table 2.34, the number of elements and
the diameters have been compiled for a general case and for the case with
d = 2.3 mm respectively, whereby the variables have the following meaning:
n: Layer number
z: Number of elements per position
6z: Total number of the elements to the layer n
d: Diameter of the cable unit
Dm: average diameter of the unit
D: Diameter of the layer
d = 2.3mm
Dm1
D1
Dm2 Dm2
D2 D2
Dm3 Dm3
D3 D3
Fig. 2.230: SI-POF cable (layer structure)
Table 2.33: Dimensions of layer-stranded POF cables in general

n z Dm D 6z n z Dm D 6z
1. 1 - 1·d 1 1. 2 1·d 2·d 2
2. 6 2·d 3·d 7 2. 8 3·d 4·d 10
3. 12 4·d 5·d 19 3. 14 5·d 6·d 24
4. 18 6·d 7·d 37 4. 20 7·d 8·d 44
5. 24 8·d 9·d 61 5. 26 9·d 10 · d 70
6. 30 10 · d 11 · d 91 6. 32 11 · d 12 · d 102
Table 2.34: Dimensions of layer-stranded POF cables with d = 2.3 mm

n z Dm D 6z n z Dm D 6z
1. 1 - 2.3 mm 1 1. 2 2.3 mm 4.6 mm 2
2. 6 4.6 mm 6.9 mm 7 2. 8 6.9 mm 9.2 mm 10
3. 12 9.2 mm 11.5 mm 19 3. 14 11.5 mm 13.8 mm 24
4. 18 13.8 mm 16.1 mm 37 4. 20 16.1 mm 18.4 mm 44
5. 24 18.4 mm 20.7 mm 61 5. 26 20.7 mm 23.0 mm 70
6. 30 23.0 mm 25.3 mm 91 6. 32 25.3 mm 27.6 mm 102
Cable Materials
The specification profile for SI-POF cable or SI-POF lines in various fields of
applications such as in industry, in office environments or in the automotive field
place the highest demands on the material components.
Thermoplastic materials (polymers) are preferred that have been mounted to the
cable using an extrusion process. Excellent mechanical properties are needed so
that the values listed below are ensured when SI-POF cable or SI-POF lines are
installed.
¾Abrasion
¾Repeating bending characteristics
¾Torsion
¾Acceleration
¾Hammer blow
¾Small bending radii
Especially in the automotive field, the material must be highly resistant to the
following properties:
¾Resistance to oil
¾Cooling lubricant resistance
¾Steam
¾Hot gases
The demand for materials that are temperature resistant comes from users.
These customers are in the automotive field, in industry or in the cable-installation
field for buildings. Special halogen-free material properties are desired in order to
provide on-site safety to customers and consumers alike.
Today’s selection of modern plastic insulation and cladding mixtures, which in
part can be improved through various methods of crosslinking, should and must
protect the SI-POF cables or SI-POF lines in all types of applications.
In case of an accident, special plastic optical fiber cables are to have emergency
running properties. SI-POF hybrid cable constructions ensure this reliability to a
very high degree.
The mechanical properties of thermoplastic materials such as
¾Hardness
¾Density
¾Tensile strength
¾Elongation at break
¾Tensile stress value
¾Compression strain
¾Impact resistance
¾Electrical properties
can be found in the relevant data specifications of the standardized norms or the
data specifications of the chemical industry. Preferred plastic materials are:
¾Polyethylene
¾Polypropylene
¾Polyurethane
¾Cross-linked thermoplastics
The properties that have been improved by cross-linking are those of thermal
resistance and higher mechanical strength. In addition, the resistance to solvents
has also been increased, which can be seen by the fact that less swelling and
cracking occur for polymers with residual tensile stress.
The essential physical properties of some of the important materials are listed
in section 3.3.6.
A very good alternative is a combination of plastic and metal, for example, with
the corrugated micro tube. Metal in the most varied constructions, whether as a
steel alloy, in aluminum or in copper keeps the SI-POF in an expanded tempera-
ture range protected against mechanical and thermal strain.
2.8.3.5 Corrugated Micro Tube Cables

Corrugated micro tubes have been used to protect cables for quite some time.
Nexans was the first company to encase polymer optical fiber wires for manufac-
turing resistant cables. Because of the small diameter of the POF, special corruga-
ted micro tubes (CMT) were needed. More detailed descriptions of the mechanical
and thermal properties are found in [Schei98], [Zam99], [Ziem99a], [Ziem99b]
and [Zam00a]. Figure 2.231 illustrates a POF wire with aluminum corrugated
tube.
Fig. 2.231: POF wires with corrugated micro tubes
Possible applications for CMT cables will be discussed later in Chapter 8.1.1.7.
The manufacturing process for corrugated tubes is described below.
Corrugated Tube Process

The UNIWEMA (Universal Corrugated Tube Machine) has become a standard
piece of equipment for cable plants worldwide. The origins of the corrugated tube
process go back to the 1940’s.
The corrugated tube process as practiced today is a butt-welding process for
small dimensions (for example POF wires). A thin metal strip is formed around a
cable core and formed into a small metal tube. The strip edges that form a butt
joint are welded into a tube cladding by a laser beam under protective gas (argon
and/or helium) and then corrugated in a spiral-shaped way or as rings (Fig. 2.232).
Fig. 2.232: Corrugated tube for POF
The UNIWEMA is used to weld copper, aluminum and steel strips or steel
alloys or alternative materials. The machine creates smooth and corrugated metal
tubing in an economical manner.
The tube welding process is continuous and fast. All weldable metals such as
copper, aluminum, steel and their alloys can be processed. The process can be
used for manufacturing small metal tubes for core diameters ranging from 1 mm
to 500 mm. Strip thickness’ of 0.05 mm to 4.0 mm are welded with a laser using
the WIG process. Neither burrs nor bulges are produced at the welding seam
(Fig. 2.233).
Fig. 2.233: Welding seams in laser welding

Due to the concentrated thermal effect of the welding source, the welding zone
is limited on the metal edges. The heat is quickly dissipated over the tube. Since
the welding zone is covered by a protective gas shield, the formation of an oxide
layer is prevented.
Corrugated Tube for POF Applications

Metal tubes manufacture in compliance with the UNIWEMA procedure
(Fig. 2.234) can be used for all POF cables. This applies to metal tubes of steel
and welded special steel alloys as well as smooth or corrugated copper or alumi-
num tubes that have been welded lengthwise. Corrugated copper tubes are used
wherever a particularly high conductivity or large dissipation of heat is required.
Due to its comparably small weight when used with thin metal strips, corrugated
cable tubes can be easily transported and installed. The corrugated tubing is easy
to bend and particularly resistant to external deformation in the radial direction. It
is absolutely hermetic. This makes it possible to operate corrugated cable tubes,
and POF elements under pressure and in a vacuum.
Fig. 2.234: Laser welding device ([LZH01])
Laser Welding
The laser beam is monochromatic and coherent and can be easily focused. As a
result, a high power density can be achieved at the processing point - the V-seam
between the strip edges (Fig. 2.235).
Fig. 2.235: Welding seam with laser beam ([LZH01])
By applying the auxiliary gases argon and/or helium in such a way that the
beam power is absorbed in the capillaries, the coupling properties of the plasma
can be controlled. The actual welding joint is produced by the melt converging
behind the capillaries (Fig. 2.236).
laser beam laser beam
metal vapour
laser induced
plasma welding zone
(fluid)
welding zone
(fluid) vapour
welding zone
(plasma) (solid)
welding zone
channel
(solid)
conduction limited welding direction of welding keyhole welding
Fig. 2.236: Principle of laser welding

Keyhole welding causes the process heat to be uniformly distributed at mini-

mum levels over the entire welding zone (Fig. 2.237). Typical welding joints are
butt-welded or overlapping welding seams, weldable materials steel, special steel,
brass, copper, aluminum and special metal alloys. Thin clad metal foils made of
aluminum/plastic/aluminum can be used for laser welding. Fluted steel sheets can
be welded overlapping or butt-jointed with a YAG laser or a diode laser.
ND:YAG-laser laser
data
beam source control controller
acquisation
device
process computer
laser fiber
quotient
pyrometer
partially
transparent optic
O reflects
Nd:YAG
radiation
laser optics O transmits heat
radiation
beam-material
interaction zone detected heat
radiation
modified track laser beam
workpiece
feed direction
Fig. 2.237: Structure of a laser welding system

2.9 Microstructured Fibers 215
2.9 Microstructured Fibers
In addition to classic optical fibers which consist of a core and cladding there are
also microstructured fibers in which the wave guiding does not rely on a refractive
index profile, but on holes along the entire length of the fiber. Normally, wave
guiding in optical fibers is based on the effect of total reflection in the general
sense of the term. The core consists of a material with a higher refractive index
than the surrounding cladding material. In this fiber configuration special field dis-
tributions, so-called modes or eigenmodes, can be guided within the fiber. These
modes experience an effective refractive index of the fiber, which lies between the
maximum refractive index of the core and that of the cladding material.
In 1996, J. Knight et. al. demonstrated a new kind of optical fiber, the wave
guide characteristics of which were no longer based on a rotation-symmetrical re-
fractive index. This created a variety of completely new possibilities and novel
functions ([Kni96] and [Kni97]). These fibers now only consist of one material,
usually silica glass, and have a structure of the cross-section with air holes. The
holes in this structure are as a rule considerably smaller than the wavelength of
light so that they do not act like objects on which light is reflected or scattered.
Instead they change the refraction characteristics of the material.
The material is changed in such a way that it acquires new kinds of charac-
teristics. Relatively simple and specific characteristics can be created with these
fibers, e.g. for dispersion, dispersion slope, modal field radius and others.
For some years now microstructured fibers have also been made of polymer.
These fibers with low temperature processes can be produced on the basis of the
low melting point of polymers and other characteristics, thus resulting in possibi-
lities for new kinds of fiber geometries and also potentially new applications.
In the following section we would like to deal with the fundamental wave
guiding mechanisms. The different types of fibers and their specific characteristics
will be introduced and the methods for producing these different types of fibers
will then be shown. We would particularly like to take a close look at the diffe-
rences between microstructured fibers made of glass and polymers. Applications
which are possible with these fibers and are presently the subject of research will
then be introduced. Some of these applications can even be obtained commercially
now. Finally, the present state of development will be discussed and we will ven-
ture a prognosis as to where the limits for such fibers may lie in the future.
2.9.1 Kinds of Wave Guiding
Wave guiding in microstructured fibers is determined by the structure of the cross-

section along the entire fiber. Holes which locally vary the refractive index very
strongly are normally put into the fiber along its entire length. These areas with
noticeably different refractive indices are very small in relation to the wavelength
so that they cannot be resolved by light and only have an indirect influence on the
propagation characteristics of the light.
216 2.9 Microstructured Fibers
There are two fundamental mechanisms which exercise this influence: holes
either act as a kind of doping by changing the effective refractive index of the
material in average ([Gho99]) or they are put into a regular, grid-shaped arrange-
ment so that they act like a kind of meta-material ([Cre99]. Other materials which
have a greatly differing refractive index from that of the core material can also be
used). Such fibers can exhibit effects with a great degree of wavelength depen-
dence since such arrangements have similar characteristics as e.g. Bragg gratings,
in which the light at certain wavelengths can be constructive or destructive over-
lapped. The two-dimensional pendant to such a Bragg grating are the Bragg fibers
in which concentric areas with greatly differing refractive indices alternate at
regular intervals ([Yeh78]). Constructive overlapping waves can come about at
certain wavelengths thus resulting in wave guiding. At other wavelengths light is
not guided. One can therefore surmise that such fibers are capable of having
strong wavelength-dependent characteristics.
A new kind of wave guiding occurs in such fibers with regular structures. This
wave guiding is possible in cores made of air as opposed to those fibers based on
total internal reflection. For wave guiding with total internal reflection it is essen-
tial that the core material has an effective refractive index which is higher than
that of the cladding. This is not necessary with fibers having a “photonic band
gap”. Because of the regular structure within the fiber band structures are formed
analogous to electrical semiconductors in which certain energy states of light
waves are allowed and others are rejected resulting in light waves which can
remain within the material and others which cannot. When there are light waves
which have permissible energy states within the core area, but not in the cladding,
then the light must stay in the core and is guided through this band gap since they
cannot exit into the cladding.
2.9.1.1 Effective Refractive Index

Fibers based on the effect of an effective refractive index can intuitively be under-
stood most easily. They are doped with the material by introducing air or other
materials. The holes made must be very small in relation to the wavelength and
should be as randomly arranged as possible. The effective refractive index then
results from the volume ratio of the two materials (e.g. one talks about air frac-
tion). The greater the proportion of air, the smaller is the effective refractive index
of the material. Fibers based on an effective refractive index should have holes
relatively small in relation to the wavelength of the light so that the holes as such
can no longer be resolved. Also, these holes should be introduced into the material
in an irregular manner as possible so that the geometry and arrangement of the
holes do not have any influence on the characteristics of the material (see
Fig. 2.228).
Such fibers are basically not different from traditional fibers in which the core
has a higher refractive index than the cladding. Consequently, there is a form of
total reflection. Since the effective refractive index can fundamentally only be
reduced by doping with air, the cladding area in such fibers is normally structured
with holes. The core is mostly undoped glass. Such fibers can be described as
being similar to normal step index glass fibers, whereby the fiber parameter de-
pends on the wavelength of the light ([Mor03a ] and [Mor05b]). The reason for
this is that the influence of the holes varies greatly depending on the wavelength
of the light which also depends on the relationship between the hole diameter and
the wavelength and whether light can resolve the holes.
Fig. 2.238: MPOF with effective refractive index according to [Lar02a]
2.9.1.2 Photonic Band Gaps

In addition to the fibers whose refractive index profile arises from the effective
refractive index resulting from the holes there is also wave guiding on the basis of
a photonic band gap ([Cre99]). Fibers based on the principle of a photonic band
gap behave fundamentally differently from the fibers with an effective refractive
index just discussed.
These fibers must have holes introduced in a specific periodic arrangement so
that a kind of meta-crystal comes about. According to the Bloch theorem the
neighboring holes act like elementary cells which are repeated regularly in several
dimensions resulting in new kinds of characteristics for the meta-crystal. Just as
with semiconductors, energy bands can be formed which originate from the perio-
dic structure of the material. In semiconductors these are the periodically arranged
atoms of the semiconductor material; in fibers with a photonic band gap it is the
periodically arranged holes.
In such fibers the light guiding comes about when the light of a certain wave-
length, and thus photons with a specific energy, possess allowed energy states in
the core area while the same energy states are not permitted in the cladding area.
Thus, the photons in this energy state can only stay in the core area of the fiber
(see Fig. 2.239).
nSiO2
1.4
effective index neff

1.3
1.2
1.1
n air
1.0
PBGF-mode
0 1 2 3 4 5 6 7
normalized frequency /O
Fig. 2.239: Distribution of intensity in a large-mode-area-laser fiber according to [Lim03]

(left). Effective refractive index of the radiation modes of the cladding (grey)
with position of the bound defect mode in the band gap (center) and the mag-
netic field strength of the linear polarized fundamental mode for ȁ = 2.27 μm,
d = 1.993 μm, D = 4.54 μm at Ȝ = 1,55 μm with neff = 0.977 (right).
The form of the energy bands, i.e. the energy areas, which correspond to the
permissible energy states is greatly dependent on the arrangement of the indivi-
dual holes. Even small deviations can lead to great changes in the energy bands so
that with this kind of fiber only slight tolerances are allowed in the arrangement of
the holes. Nevertheless, these fibers permit greater possibilities for structuring
([Arg06]). As a consequence, propagation characteristics such as dispersion, dis-
persion slope, effective area, etc. can have relatively large dimensions. Especially
for very narrow-band applications, e.g. sharp-edged filters, fibers with photonic
band gaps can be employed quite well. This is also true for high-performance
applications in which the linear characteristics of the hole core are used ([Lim03],
[Mat05b] and [Nie06]).
Fig. 2.240: Air-hole - MPOF with 220 μm outer diameter/5 μm hole distance, [Eij03a]
2.9.1.3 Bragg Fibers

Bragg fibers consist of concentric rings with different refractive indices. These
rings act like a Bragg grating in radius direction so that they reflect certain wave-
lengths which are adapted to the spacing between the rings while letting other
wavelengths through. This results in wave guiding in only those wavelengths
which the Bragg rings reflect. With all other wavelengths no wave guiding takes
place. The fibers thus act like a filter and only let light through with very specific
wavelengths ([Yeh78]).
Fig. 2.241: Cross section of a Bragg-fiber according to [Arg06]
The rings can be produced in a variety of different ways. Refractive index pro-
files can be produced which have higher or lower refractive indices with specific
radii. Microstructured fibers, however, are also possible where the rings with dif-
ferent refractive indices are realized by hole structures. In this case rings with
holes are arranged at regular distances from the fiber axis which, because of the
effective index of this layer, acts like a layer with reduced refractive index.
Bragg fibers behave similarly to fibers with photonic band gap. They are also
based on the exact arrangement of the holes or the layers with different refractive
indices respectively. If the geometry is followed exactly very sharp-edged filters
can be produced or fibers which are very selective in regard to the wavelength.
2.9.1.4 Hole-Assisted Fibers

In addition to these new kinds of fibers whose waveguide characteristics are based
solely on the structures introduced, hybrid fibers have also been introduced which
represent a cross between conventional fibers with refractive index profiles and
microstructured fibers ([Has01]). These fibers have the same wave guiding as with
conventional fibers. However, the additional holes change the propagation cha-
racteristics so that you get other degrees of freedom in fiber design. In particular
ring-shaped hole structures are arranged around the core in order to reduce the
bending sensitivity of the fibers ([Guan04] and [Nak03b]). The outer structure acts
like an additional step in the refractive index profile which should hold part of the
output emitted in the bend in the cladding area. This measure is supposed to
increase the wave guidance without having to make compromises concerning the
propagation characteristics of the fiber.
Fig. 2.242: Cross section of a hole-assisted fiber according to [Guan04]
2.9.2 Production Methods
Microstructured fibers can be produced in very different ways. Various production

methods are possible with glass and polymer fibers.
2.9.2.1 Microstructured Glass Fibers

The first microstructured fibers were made of glass ([Kni96] and [Kni97]). Since
glass has a very high melting point the production possibilities are limited. The
fibers are mostly produced using the so-called stack-and-draw technique in which
small glass tubes with different diameters - the number depends on the desired
hole diameter - are put together in a bundle. Depending on the type of fiber, either
a filled glass rod (effective index) or another small glass tube (photonic band gap)
is used for the core. These small tubes combined then form the preform. They are
melted and drawn into a fiber. A fiber cladding is generally drawn over the entire
preform which then forms the outer area of the fiber. This only serves the purpose
of stabilizing the fiber.
The fact that the small round glass tubes are combined into a preform generally
only allows a few arrangements: rectangular, hexagonal or so-called honey comb
structures. Even if you decide on hexagonal structures when arranging the holes,
the hole spacing and the hole sizes can be put into a rather large range which can
lead to diverse design possibilities.
Fig. 2.243: Cross section of a microstructured glass fiber, fabricated by stack-and-draw-

technology ([Ort04])
Glass melting at low temperatures can also be extruded, whereby the glass is
either melted or liquefied. The ensuing viscous fluid can then be pressed through
specifically arranged nozzles which have the structure of the desired preform. This
preform can then be used to immediately draw the fiber or to make a preform.
This method of making preforms in effect allows the production of as many hole
geometries as one likes. In principle, round holes and any kind of arrangement can
be produced in this way. However, this production method is limited to glass with
a low melting point. Consequently, silica glass, for example, cannot be processed.
The production engineering of microstructured fibers has improved tremen-
dously in the past few years. Whereas the first fibers still had attenuations of seve-
ral 100 dB/km, today fibers based on an effective index with attenuations per unit
length can be produced below 0.3 dB/km at a wavelength of 1.55 μm ([Taj03]).
Photonic band gap fibers permit attenuations per unit length up to 13 dB/km
([Smi03]).
2.9.2.2 Microstructured Polymer Fibers (MPOF)

Fibers made of plastics can be produced in a variety of different ways, especially
since they can be processed at much lower temperatures. Whereas glass fibers can
be drawn at temperatures around 2000°C, MPOF can already be drawn at 200°C
([Lyy04]). This not only allows simpler production techniques, but also permits
the introduction of other materials into the fiber which would otherwise decom-
pose, e.g. dyes ([Lar04]). However, there are also some disadvantages in regard to
increased attenuation, lower operating temperatures, other operating wavelengths,
etc. ([Lar06a]).
Microstructured polymer fibers can also be extruded and then drawn into fibers.
The same limitations regarding geometry and production tolerances are valid for
them as for glass fibers.
Researchers at the University of Sydney ([Bar04c] and [Lar01b]) have
developed a particular kind of preform production in which a massive cylinder
made of polymer is structured using drills with different diameters. At present,
preforms up 65 mm in length can be structured with this method, otherwise the
drills would be too long. As many geometries as one may wish can be produced in
which both the arrangement and the hole diameter can be freely chosen. Present-
day production processes have hole diameters between 1 mm and 10 mm with
minimum spacing in between of about 100 μm which then shrink to their original
size through drawing.
New kinds of process techniques can even produce elliptical holes which give
the fiber an intrinsic double refraction. Preforms can either be poured into molds
or around capillary tubes and then drawn into fibers ([Zha06]).
Fig. 2.244: Preforms of MPOF ([Lwin06], [Poi06e])
Other materials can be introduced into the fiber in addition to the holes. Fibers
with metal wires for the poling of the material have been demonstrated as well as
fibers with liquids in the capillary for controlling the propagation characteristics
and doping materials for changing the optical and electrical characteristics
([Cox03b] and [Cox06]).
After the first MPOF was introduced at the end of 2001 ([Lar01b]), the tech-
nology has continued to develop at an amazing pace. The fibers introduced back
then still had an attenuation of 30,000 dB/km. In the course of time the individual
process parameters have been continuously improved so that the attenuation could
be steadily reduced. The process parameters optimized include conditions when
drilling the preforms, rinsing and cleaning steps as well as drawing parameters.
The best microstructured polymer fibers today have an attenuation of 200 dB/km
and are thus not very far away from conventional polymer fibers which have an
attenuation of about 120 dB/km at a wavelength of 650 nm.
100
achieved attenuation [dB/m]
Sept. 2001
10
April 2005
0.1
0 5 10 15 20 25 30 35 40 45
months for the first publications
Fig. 2.245: Development of the attenuation of the MPOF 2001-2005 ([Lwin05])
2.9.2.3 End Surface Preparation

Microtome cutting has proved to be a useful method for working on the end sur-
faces of conventional polymer fibers. This method of work only produces unsatis-
factory results with microstructured fibers since the fine, step-like structure at the
end of the fiber in the holes can lead to defects and irregularities (see Fig. 2.246).
These structures can be seen at the ends of all such fibers and on conventional
polymer fibers, too. Nevertheless, it can be seen that the mechanical characteris-
tics in particular of the MPOF intensify the step-like effect. These filigree struc-
tures absorb the lateral forces and give in again after each thrust.
Fig. 2.246: Singlemode-MPOF cut by microtome, 1000-fold magnification
The direct cutting of the fiber with conventional cutting pliers can destroy the
fine structures because of these lateral forces.
Fig. 2.247: Singlemode-MPOF cut by MOST®-tongs, 100-fold magnification
Other processing methods such as hot plate or subsequent polishing have also
been investigated, but did not deliver any good results. The hot plate technique
leads to inclusions at the end surfaces so that the original geometry can no longer
be recognized. On the other hand, polishing leads to the deposition of rubbed off
shavings and their removal into the holes. A reproducible coupling is therefore not
possible since the influence of these inclusions or that of the deposited foreign
matter in the structure’s holes is not controllable.
Better processing characteristics are shown by those MPOF which are surroun-
ded by another, so-called buffer layer made of hard polyester. This layer absorbs a
large part of the mechanical forces when cutting and prevents the breaking of the
fine webs within the structure. Since such fibers consist almost exclusively of
polymer they can almost be worked on like polymer fibers. Figure 2.248 shows
the end surface of such an embedded fiber with a buffer layer. You can see that
the fiber is not embedded centrically which leads in practical use to a lateral mis-
alignment of the plugs and thus to plug losses and power redistribution. In the
future you can expect, however, that the dimensions of the fibers will become
greater and that the fibers can be better centered with new drawing techniques.
Fig. 2.248: End face of an embedded fiber with buffer layer; 100-fold magnification
No practical solution exists yet which can provide for good reproducibility and
a high degree of reliability. Processing methods still have to be found for both
practical and laboratory use which can meet the necessary requirements. In the
case of termination in the field the end faces must allow acceptable losses; in the
laboratory, preparation with high reproducibility is necessary. Both kinds of pre-
paration still have to be developed.
2.9.3 Applications for Microstructured Fibers
Microstructured fibers allow a number of applications since their characteristics

can be adapted to wide areas as desired because of the additional degree of free-
dom in design and production. For example, waveguide characteristics such as
chromatic dispersion and its slope can be adjusted as well as the mode field dia-
meter. Other materials or fluids can be introduced into the fiber through the holes
running along the fiber. These materials can change the propagation charac-
teristics through which tunable components or sensors are made possible. Some
possible applications for microstructured fibers are subsequently described. This
list does not make any claim to being complete, but is solely intended to present
the best-known applications as well as the commercial applications available
today.
2.9.3.1 Dispersion Compensation

The first applications for microstructured glass fibers were the compensation for
dispersion or its slope respectively. Because of the additional possibilities for fiber
design, selection of the number of holes, their size and distance from one another,
wavelength-dependent effects in particular such as chromatic dispersion can be
adjusted very well. As already described above, the holes have weaker wave gui-
ding at small wavelengths because the light can enter the bridges between the
holes. This causes a different kind of wave guiding so that the fiber behaves as if it
had another fiber parameter. By skillfully selecting the diameter of the holes and
their spacing, the dispersion and higher orders can be adjusted very well. Disper-
sion-compensating microstructured glass fibers are commercially available today.
2.9.3.2 Endlessly Singlemode

Microstructured fibers also allow applications which are not possible with conven-
tional fibers. Such an application are the so-called endlessly singlemode fibers
which have one mode in the entire wavelength spectrum and do not have a cut-off
frequency. This characteristic can come about when the wave guiding changes
with the wavelength.
In step index fibers the existence of one mode is clearly determined by the fiber
parameter V which is proportional to the core diameter, the numerical aperture
and the reciprocal value of the wavelength used. Thick fibers with large numerical
apertures are characterized by a large fiber parameter V. Fibers are only guide
only one mode for V < 2.405, the first zero of the Bessel function of zeroth order.
If the wavelength selected is large enough then V will become small enough at
some point so that the fiber becomes singlemode. In microstructured fibers the
fiber parameter is not simply anti-proportional to the wavelength since the holes in
the cladding area act differently with large wavelengths than with small ones
leading to a wavelength-dependent numerical aperture so that fibers can be pro-
duced which are singlemode for all wavelengths ([Bir97], [Mor03b] and [Zag04]).
2.9.3.3 Birefringence
Since microstructured fibers are not rotation-symmetrical such as conventional fi-
bers with a refractive index profile, for example, they tend to be birefringent.
Typical hexagonal structures do not exhibit any birefringence. However, when this
symmetry is disrupted, e.g. through production tolerances, then these fibers are
birefringent.
This effect is used positively in some fibers, whereby the high birefringence
causes the fibers to retain their polarization ([Ort04]). In the case of very great
differences between the propagation constants of both polarizations they can then
only very weakly interact with each other and exchange power. When only one
polarization is launched into the fiber, then the power in this polarization is
retained and is propagated in this way to the end of the fiber.
Fig. 2.249: High birefringent MPOF by incorporated asymmetry ([Issa04b])
The effect of birefringence can be generated in microstructured fibers in two

ways: either the holes are arranged asymmetrically so that a geometric birefrin-
gence occurs which can be created in a controlled and thermally stable manner, or
the holes are elliptical and not round which contributes to the birefringence
([Issa04b]). It is more difficult to control this kind of birefringence, but it does
allow complete freedom of fiber design because the arrangement of the holes and
their size can be freely chosen.
2.9.3.4 Highly Nonlinear Fibers

The nonlinear characteristics of fibers are influenced on the one hand by the
nonlinearity of the material and on the other by the level of confinement which is
described by the so-called effective mode area. With very strong wave guiding,
light is guided into the center of the core and the optical power can propagate in
the area of the core-cladding interface layer or even in the cladding. Here the light
is strongly concentrated in a small area of the core which results in very high
intensities with the same power which can lead to nonlinear behavior within the
fiber. Such strong wave guiding can only be attained by means of big differences
in the refractive index between the core and cladding.
In conventional fibers the differences in refractive index are in the range of a
few percentage points. Microstructured fibers on the other hand consist of areas of
glass with a refractive index of about nglass | 1.5 and holes, which are generally of
air (nair | 1). The very high confinement can be achieved by this very high contrast
in refractive index.
In fibers based on this effective refractive index, the cladding has to have a very
high air fraction. The proportion of air in relation to the entire volume has to be so
high that the effective refractive index lies near the value for air. Fibers with effec-
tive area up to Aeff | 2.85 μm2 have been realized using this process ([Lee02]).
In addition, other materials such as Bi2O3 can be used which have a highly non-
linear susceptibility Ȥ3. With such materials nonlinear parameters of
Ȗ = 1100 W-1km-1 can be produced ([Lee06c]).
2.9.3.5 Control of the Effective Area

Fibers with a particularly high nonlinearity are needed for all optical signal pro-
cessing. There are, however, a number of applications in which the nonlinear
effects should be particularly weak so that the light propagation in such fibers is
not disrupted. In such fibers the opposite path is taken as with highly nonlinear
fibers: the material used should be as slightly nonlinear as possible and the effec-
tive area of the fiber should be as large as possible so that the intensity within the
fiber remains low at the given luminous efficiency. Even if the difference in
refractive index between the core material and the holes continues to remain large
you can still see to it through skillful fiber design that the light is guided relatively
weakly and the mode field takes up as large an area as possible.
In general, these fibers have a very low air-fill factor so that the effective refrac-
tive index in the cladding area lies only slightly below that of the core. Fibers with
effective areas of Aeff | 100 μm2 have been presented by [Kim06c] and [Sai06].
This technique can also be used for controlling the form of the mode field in
order to adapt it to other types of fibers and thus minimize coupling losses at the
connector. For example, Furukawa introduced such fibers at the ECOC in 2004
([Guan04]) the mode fields of which are adapted to standard singlemode fibers.
2.9.3.6 Filters
Microstructured fibers can show very strong wavelength-dependent effects. As
described above, the dispersion can be adapted to a wide area, but other wave-
length-dependent characteristics can be specifically designed, e.g. group velocity
or even the attenuation per unit length of the fiber.
Fibers with an effective refractive index permit the relatively simple adaptation
of the group velocity with which one can generate all-pass filters with specific
phase responses.
Fibers based on a photonic band gap can have very sharply delimited wave-
length ranges with which light is guided. Thus, filters with specific amplitude
response and sharp edges can be produced ([Vill03], [Kim05c], [Kim06d] and
[Sai05]).
2.9.3.7 Sensor Technology, Tunable Elements

The characteristics of microstructured fibers can be manipulated in many ways. In
particular materials can be introduced into the holes along the fiber which can
change the characteristics of the microstructured fiber through their different re-
fractive indices. These materials can be gases or liquids which are guided through
the fiber and can alter the characteristics when the composition is changed
([Car06b]).
With such methods you can also analyze liquids such as blood in the human
body. Polymer fibers are especially attractive for this kind of application because
glass can split and would thus be considered too dangerous in the human body.
You can also intentionally change the characteristics by means of the controlled
introduction of liquids. Thus, sensors have been introduced which are based exact-
ly on this phenomenon, e.g. a liquid is pushed into the capillaries in the cladding
area when the temperature rises, thereby changing the propagation characteristics
of the fiber ([Jen05]).
Consequently, the dispersion ([Gun06]) or the band gap ([Sun06]) can be
adjusted to a lesser or greater extent by introducing liquids.
Pressure sensors represent another application. Since the geometry of the holes
has a great influence on the fiber’s propagation characteristics, lateral pressures
can have a very noticeable effect on its behavior ([Eij03b]).
Especially fibers based on a photonic band gap react very sensitively to
changes in the geometry. As a result, microstructured fibers can be produced
which work like filters, the passband of which is changed when pressure is
exerted.
2.9.3.8 Double-Core and Multi-Core Fibers

Most microstructured fibers consist of a cladding area in which the holes are
arranged symmetrically or asymmetrically. In fibers based on an effective refrac-
tive index the core consists of an area in which a hole has been left out of the
arrangement. The core is thus a kind of imperfection within the photonic crystal.
In this way two or more cores can be produced by introducing two or more imper-
fections within the cladding area in which the light can propagate instead of
having just one hole in the middle. Each individual location where a hole has been
left out and the core material exists can be viewed as a separate fiber in which
light can be propagated. If the individual cores are placed far enough apart, they
either do not influence each other at all or only slightly.
Such fibers with several cores can be used for parallel data transmission
([Eij06a]). The arrangement of the individual cores is retained and so these fibers
can be used like a well-ordered fiber bundle. However, these fibers have a consi-
derably smaller diameter and can be laid like individual fibers ([Eij03b] and
[Pad04]).
Fig. 2.250: Double core-MPOF with 9.6 μm spacing between the cores ([Eij03b])
2.9.3.9 Imaging
As we have seen above, microstructured fibers can be produced with more than
one core for parallel data transmission. If you continue to increase the number of
cores, you can use the same method to produce image guides in which every indi-
vidual core transmits a part of the image (a pixel). As mentioned above, the
arrangement of the holes stays the same and the cores along the fiber are retained.
Each individual pixel reaches the end of the fiber in its definite position so that the
image is retained ([Eij04c]).
Fig. 2.251: Image guide-MPOF ([Eij04c])
2.9.3.10 Multimode Graded Index Fibers

The fibers introduced so far are relatively thin singlemode fibers. In addition to
these fibers, graded index multimode fibers made of polymer have also been deve-
loped, so-called GI-MPOF ([Kle03b] and [Eij04d]), which have the large core dia-
meter of a polymer fiber and the effective graded index profile of a multimode
glass fiber (see Fig. 2.252).
Fig. 2.252: Schematic cross section of a GI-MPOF ([Kle04b] and [Lwin06])

Polymer fibers offer a number of advantages, especially with fibers having

large core diameters compared to glass fibers (these advantages are also valid for
other types of fibers). These considerably larger core diameters are possible with-
out the fiber becoming inflexible. For this reason graded index polymer fibers
have been produced for some years now which attain core diameters into the milli-
meter range. However, these fibers have a refractive index profile in the core
which has been adjusted through doping and they are quite difficult to produce
when the core diameters are very large. Another advantage of these microstruc-
tured fibers is the lack of doping material which results in these fibers having very
good thermal and aging stability of the profile. Graded index profile polymer
fibers which already exist are not particularly thermo-stable. With aging and
especially in combination with increased temperatures they exhibit a flattening of
the profile through diffusion of the doping material. This leads to an alignment of
the concentrations of the doping materials resulting in a leveling out of the profile.
Fig. 2.253: Cross section of a graded index profile multimode polymer fiber (GI-MPOF)
with 135 μm core- and 520 μm outer diameter ([Eij04d]) and of a MPOF
according to [Lwin06]
Figure 2.253 shows a multimode fiber in which the effective refractive index
continuously decreases with increasing distance to the fiber axis. If you take an
average of the entire circumference of the refractive index, then you have a para-
bolic refractive index profile in the radius direction. Measurements have shown
that these fibers have a similar propagation behavior as a conventional multimode
fiber. However, the differences lie in the detail. If you stimulate the GI-MPOF
with a small spot, for example, the fiber behaves differently, depending on
whether or not the light hits a hole or the core material; something that cannot
happen in conventional fibers. For this reason greater research and development in
measurement techniques and characterization are necessary before the GI-MPOF
is widely used in commercial applications.

Fundamentals of Optical Fibre

Uploaded by

Document Information

Original Description:

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

Fundamentals of Optical Fibre

Uploaded by

Copyright:

Available Formats

2.

2.1 Fundamentals of Optical Fibers

2.1.1 Refractive Index Profiles

Fig. 2.1: Refractive index profile in a step index profile fiber

where a is the core radius.

where g is the profile exponent

2.1.2 Numerical Aperture

n0  sin 4max ncore 2  ncladding 2 , for n0 1 follows

sin 4max ncore 2  ncladdimg 2

AN sin 4max ncore  2  '

Fig. 2.3: Definition of the acceptance angle

Fig. 2.4: Acceptance angle of a graded-index profile fiber

singlemode glass fiber multimode glass fiber polymer fiber

multimode glass fiber 62.5/ 125 μm

2.1.3 Ray Trajectory in Optical Fibers

Fig. 2.6: Meridional rays

Fig. 2.7: Skew rays in step index profile fibers

2.1.4 Modes in Optical Fibers

2.1.4.1 The Mode Concept

O/ncore transverse direction

Fig. 2.9: Formation of the mode structure within the waveguide

Fig. 2.10: Guided, cladding and radiation modes

2.1.4.2 Mode Propagation in Real Fibers

only a few launched discrete modes

nearfield (very schematically)

2.1.5 Parameters for Describing Real Fibers and Waveguides

In order to describe the characteristics of real fibers and waveguides different

Fig. 2.14: Definition of attenuation

Attenuation value a is the non-dimensional variable (given as a number or in

attenuation factor a [dB]

Quantity Symbol Unit Formula

Especially in the area of optical short-range communication, indicating the fiber

Nevertheless, the representation comprises approximately 3 decades, i.e. a fac-

2.1.5.2 Mode-Dependent Attenuation

Fig. 2.17: Goos-Hänchen shift

2.1.5.3 Mode Coupling

Fig. 2.18: Mode coupling at a scattering center

Fig. 2.19: Mode coupling at the core/cladding interface

Mode coupling alters the bandwidth of a fiber. When collimated light is

2.1.5.4 Mode Conversion

fiber axis behind

new propagation angle

Fig. 2.21: Mode conversion at a bend

Strictly speaking, mode conversion can be described as a special case of mode

An impressive experiment that confirmed this statement was shown in [Poi00].

2.1.5.5 Mode Coupling Lengths

Fig. 2.23: Attenuation of a SI fiber under different excitation (acc. to [Lub02b])

far field width [°] NALaunch: 0.64 0.33 0.09

Fig. 2.25: Launch dependent far field widths of a PMMA SI-POF

In general, the mode coupling length is characterized as the distance in which a

mode dependent fiber parameter [a.u.]

equilibrium mode value

2.1.5.6 Leaky Modes

2.1.5.7 Dispersion in Optical Fibers

The bandwidth of a fiber communication transmission system is usually con-

Fig. 2.28: Effect of dispersion on a sine-wave signal

Fig. 2.29: Pulse broadening by passing an optical fiber

modal dispersion chromatic dispersion

profile material dispersion

polarization mode dispersion waveguide dispersion

Fig. 2.30: Dispersion mechanisms in optical fibers

2.1.5.8 Mode Dispersion

Fig. 2.31: Deriving the difference in the transit time

n0 sin 4max ncore 2 ncladding 2 , for n0 1 follows

sin 4max ncore 2 ncladdimg 2

AN sin 4max ncore 2 '

300 core t 750 μm