1. Introduction
It is well known that a radiation loss occurs when a singlemode fiber is bent. The
bend loss has been regarded as an adverse effect in the context of optical
transmission. However, bend fibers can also be optimized as novel optical components
which can be employed in optical communications or optical sensing [
1–3
R. C. Gauthier and C. Ross, “Theoretical and experimental
considerations for a singlemode fiber-optic bend-type
sensor,” Appl. Opt.
36, 6264–6273
(1997). [CrossRef]
]. Bend loss for a standard singlemode fiber
(SMF28) has been investigated theoretically and experimentally in our previous
research and it has been indicated that the polymer coating layer, which offers the
mechanical protection, has a significant influence on the bend loss when the bending
radius is relatively small, e.g., within the range from 8 to 13 mm for the
wavelength ranged from 1500 to 1600 nm [
4
Q. Wang, G. Farrell, and T. Freir, “Theoretical and experimental
investigations of macro-bend losses for standard single mode
fibers,” Opt. Express
13,
4476–4484(2005). [CrossRef] [PubMed]
]. In Ref. [
3
Q. Wang, G. Farrell, T. Freir, G. Rajan, and P. Wang, “Low-cost wavelength measurement
based on a Macrobending Singlemode Fiber,”
Opt. Lett.
31, 1785–1787
(2006). [CrossRef] [PubMed]
] the macrobending SMF28 has been developed to act as an edge
filter through optimizing the bending length and bending radius according to the
required discrimination in bend loss over the wavelength range. Utilizing this novel
low cost edge filter, an all-fiber ratiometric wavelength measurement system has
been demonstrated.
For an edge filter employed in wavelength measurements, low polarization dependence
is also desired besides the discrimination in bend loss over the wavelength range.
However, the investigation on bend loss for SMF28 in Ref.[
4
Q. Wang, G. Farrell, and T. Freir, “Theoretical and experimental
investigations of macro-bend losses for standard single mode
fibers,” Opt. Express
13,
4476–4484(2005). [CrossRef] [PubMed]
] and the corresponding optimal design in Ref. [
3
Q. Wang, G. Farrell, T. Freir, G. Rajan, and P. Wang, “Low-cost wavelength measurement
based on a Macrobending Singlemode Fiber,”
Opt. Lett.
31, 1785–1787
(2006). [CrossRef] [PubMed]
] are based on the scalar approximation, i.e., the
polarization dependence of bend loss has not been considered yet. For weakly guiding
singlemode optical fibers, scalar approximation of the wave equation works well for
the analysis of the light propagation. Generally speaking, the difference in the
refractive index between the polymer coating layer and the cladding layer is much
higher as compared to that between the cladding and core. The reflectance of the
radiated field occurring at the interface between the coating layer and cladding
layer is believed to differ for different polarization states, which can lead to
polarization dependence of the bend loss due to the coupling between the reflected
radiated field and the quasi-guided fundamental mode. However, to the best of our
knowledge, the previously published investigations on the bend loss for the
singlemode fiber were carried out for a scalar approximation [
4–7
Q. Wang, G. Farrell, and T. Freir, “Theoretical and experimental
investigations of macro-bend losses for standard single mode
fibers,” Opt. Express
13,
4476–4484(2005). [CrossRef] [PubMed]
].
This paper presents both theoretical and experimental investigation on the
polarization sensitivity of macrobending fiber with a relatively small bending
radius. In Section 2, the calculation formulation of bend loss of the singlemode
fiber is refined for the TE and TM mode separately, which takes account of the
respective boundary condition at the interface between the cladding and the coating
layer. The normalized polarization dependent loss (PDL) is proposed for better
characterization of the polarization dependence for different bend radii. Section 3
presents the corresponding experimental tests, which show a good agreement with the
theoretical results. Both the theoretical and experimental results show that the
coating layer has a significant influence on the polarization sensitivity of bend
loss.
2. Calculation of bend loss for TE and TM mode
Fig.1. (a). Schematic structure of a bending fiber consisting of core, cladding,
coating and absorbing layer; b) squared effective index profile with
conformal mapping of the bending fiber.
Figure 1(a) presents a schematic structure of a bending fiber
consisting of a core, cladding, coating and absorbing layer. The function of the
absorbing layer is to absorb the radiated field that penetrates through the coating
layer as presented in Ref. [
3
Q. Wang, G. Farrell, T. Freir, G. Rajan, and P. Wang, “Low-cost wavelength measurement
based on a Macrobending Singlemode Fiber,”
Opt. Lett.
31, 1785–1787
(2006). [CrossRef] [PubMed]
] so that the influence of the interface between the coating
layer and air on the bend loss is removed. The structure is equivalent to a
core-cladding-infinite coating structure. For this structure, the reflection of the
radiated field occurring at the interface between the cladding and coating has a
significant effect on the bend loss when the bend radius is relatively small due to
the whispering-gallery mode caused by the interface.
Conventionally to predict the bend loss, the formulas developed in Refs. [
5–7
I. Valiente and C. Vassallo, “New formalism for bending losses in
coated single-mode optical fibers,”
Electron. Lett.
25,
1544–1545(1989). [CrossRef]
] can be employed. However, the formulas
presented in these references are based on the scalar approximation. Calculation of
bend loss taking account of the polarization of the propagation light can be based
on a rigorous three-dimensional full-vector wave equation using a cylindrical
coordinate and be solved with alternative numerical methods, such as the beam
propagation method (BPM) or finite-difference (FD) method [
8
R. Scarmozzino, A. Gopinath, R. Pregla, and S. Helfert, “Numerical techniques for modelling
guided-wave photonic devices,” IEEE J.
Sel. Top. Quantum Electron.
6, 150–162
(2000). [CrossRef]
,
9
M. S. Stern, “Semivectorial polarized finite
difference method for optical waveguides with arbitrary index
profiles,” IEE Proceedings
135, 56–63
(1988).
]. These numerical methods can be used to calculate the bend
loss for the TE and TM modes separately [TE mode: the polarization direction is
defined in the x-z plane; TM mode: the polarization direction is in the y-z
direction; see
Fig. 1(a)], but it is time consuming. According to the
related investigations on the integrated bending waveguide, a conformal mapping can
also be applied to the full-vector wave equation with a good approximation [
10
W. W. Lui, C. -L. Xu, and T. Hirono, et.al., “Full-vectorial wave
propagation in semiconductor optical bending waveguides and equivalent
straight waveguide approximations,” J.
Lightwave Technol.
15, 910–914
(1998). [CrossRef]
,
11
W. Berglund and A. Gopinath, “WKB analysis of bend losses in
optical waveguides,” J. Lightwave
Technol.
18,
1161–1166(2000). [CrossRef]
]. In addition, for these normal bending waveguide structures
(contrast to the bending waveguide based polarization rotators), the cross coupling
of transverse field is neglectable. Therefore, a quasi-vector wave equation is a
good approximation for the bending fiber. Based on the quasi-vector approximation,
the transverse field distribution in the cladding and coating area of the bend fiber
(exclude the interface between the cladding and coating) satisfies the equation [
5
I. Valiente and C. Vassallo, “New formalism for bending losses in
coated single-mode optical fibers,”
Electron. Lett.
25,
1544–1545(1989). [CrossRef]
,
6
H. Renner, “Bending losses of coated singlemode
fibers: a simple approach,” J. Lightwave
Technol.
10, 544–551
(1992). [CrossRef]
],
Where
ψ is the electric component
Ey
for the TM mode or the magnetic component
Hy
for the TE mode.
and
λ is the wavelength in free-space.
neff
is the effective index distribution of the
bending fiber and
n
2
eff
(
x,y) =
n
2 (
x,y)(1 +
2
x/
R) as shown in
Fig. 1(b).
β is the complex
propagation constant, of which the imaginary part is the bend loss coefficient
α (the bend loss can be calculated by
Ls
=10log
10(exp(2
αL))).
Reference [
6
H. Renner, “Bending losses of coated singlemode
fibers: a simple approach,” J. Lightwave
Technol.
10, 544–551
(1992). [CrossRef]
] presented a detailed derivation for solving the bend loss
coefficient based on some approximations, e.g., the curved interface between the
cladding and coating is treated as a plane and the light field within the core is
approximated by the unperturbed field of the straight fiber with infinite cladding.
Following the approximations and formulations in that reference
where .Bi
and Ai
are Airy functions, respectively.
For the TM mode, with the boundary condition between the coating layer and cladding
layer, we have
For the TE mode, the corresponding boundary condition between the coating layer and
cladding layer is
With these two different boundary conditions bend loss coefficient can be determined
as αTE
and
αTM
respectively, and consequently, we can
calculate the bend loss for the TE and TM mode respectively, namely,
LsTE
and LsTM
.
There is an anisotropic variation of the refractive index of fiber due to the
mechanical stress caused by bending. To take account of the influence of the
variation of the refractive index on fiber bend loss, a practical method in the
calculation of the fiber bend loss is to introduce an elasto optical correction for
the agreement with the experiment, i.e., using a so-called effective bending radius
Reff
and
Reff
=
(1.27 ∼ 1.31)
R
exp
(
R
exp is the actual bending radius in the experiment).
However, our previous investigation in Ref[
4
Q. Wang, G. Farrell, and T. Freir, “Theoretical and experimental
investigations of macro-bend losses for standard single mode
fibers,” Opt. Express
13,
4476–4484(2005). [CrossRef] [PubMed]
] indicates that this elasto optical correction is not
required for the SMF28 in the modeling for the considered bending radii, which
suggests that the contribution of variation of refractive index for SMF28 to the
bend loss is neglectable. Therefore, the stress effect on bend loss of the two
polarization modes is neglected in this paper.
For SMF28, the refractive index of the core, cladding and coating layer is 1.4504,
1.4447 and 1.4782, respectively at the wavelength 1.55 μm, the radius of
the core and cladding is a=4.15 μm and b=62.5 μm. For a
bending radius from 8 to 13 mm and a bending length of 10 turns, the bend loss for
TE and TM mode is calculated and presented in
Fig. 2(a). From
Fig. 2(a), it can be seen that because of the
whispering-gallery mode, the bend loss does not increase monotonically. Using the
conventional definition, the polarization dependent loss can be calculated by
PDL =
LsTE
-
LsTM
(dB) and for the bend loss presented in
Fig. 2(a), the corresponding PDL are presented in
Fig. 2(b). From this figure one can see that with the same
bending length the polarization dependent loss of a bending fiber differs
significantly from different bending radii. Similar to the bend loss, the absolute
value of polarization dependent loss does not increase monotonically as the bending
radius decreases, e.g., PDL for bending radius of 9 mm is much bigger than those of
8 and 8.5 mm.
Fig. 2. (a). Bend loss for TE and TM mode; (b). Polarization dependent loss for the
bending fiber.
For the design of bend loss edge filter comparison of polarization sensitivity for a
bending fiber for different bending radii is required regardless of the bending
length even where the actual loss values are closely matched. Choose the above SMF28
as an example and two cases are considered. Case 1) a bending length of 20 turns for
bending radius of 10 mm and case 2) a bending length of 10 turns for the bending
radius of 10.5 mm. Within the wavelength range from 1500 to 1600 nm, they have close
bend losses as presented in
Fig. 3(a), both of which are suitable to act as an edge
filter for wavelength measurement if only considering the bend loss. However, the
polarization dependent losses for these two cases plotted in
Fig. 3(b) are significantly different. From
Fig. 3(b) it can be seen that the PDL for case 1 is bigger
than that of case 2. So the polarization dependent loss with the traditional
definition is not sufficient to characterize the polarization sensitivity of bend
loss for different bending radii in a general sense, i.e., regardless detailed
bending loss and bending length.
Fig. 3. (a). Bend loss for TE and TM mode of a bending fiber with radius 10 mm and
length of 20 turns and a bending fiber with radius of 10.5 mm and length of
10 turns; (b) corresponding polarization dependent losses.
Therefore, a normalized polarization sensitivity is proposed by the following
definition
With this formula, the polarization sensitivity for a bending radius can be
characterized in a general sense without referring to a detailed bend loss or
bending length. In a practical situation, the polarization dependent loss can also
be easily estimated providing the bend loss of the fiber is known.
For the first calculation example, the polarization sensitivity for different bending
radii at a wavelength of 1550 nm is presented in
Fig. 4. Compared to the
Fig. 2(b), this normalized polarization dependence can
predict the polarization sensitivity for different bending radii accurately. For
example, in
Fig. 4, the polarization dependence for bending radius of 10
mm is much higher (here the absolute value is referred) than that for bending radius
of 10.5 mm, which has been verified in
Fig. 3(b).
Fig. 4. Normalized polarization dependent loss for different bending radii.
The normalized polarization dependence shown in
Fig. 4 shows that the polarization sensitivity of SMF28 has a
quasi-periodical characterization with bending radius and this quasi-periodical
characterization is very close to that of the bend loss vs bending radius as shown
in
Fig. 2(a). However, in
Fig. 2(a), it can be seen that although because of the
whispering gallery mode, the bend loss does not decrease monotonically as the
bending radius increases; but it still shows a decreasing trend. For example, the
bend loss for bending R=9 mm and R=10.7 mm are both on the peaks of the curve, but
the bend loss the bend loss for bending radius of 9 mm is much bigger than that of
R=10.7 mm. As distinct from the bend loss, the normalized polarization dependence in
Fig. 4 has no such trend, which means that for bending radius
R=9 mm and R=10.7 mm, they will have the same polarization dependent loss when they
have the equivalent bend loss.
To verify the polarization sensitivity of the bending fiber, corresponding
experiments are presented below.
3. Experimental results
The polarization dependence of the bending fiber was measured for different bending
radii and different wavelengths, respectively.
Figure 5 gives the experimental setup used for our
measurement of polarization dependence of fiber bend losses, which includes a
tunable laser, a polarization controller, a bending fiber section and an optical
spectrum analyzer.
Fig. 5. Experimental setup for measuring polarization dependent loss.
First, the bend loss of the SMF28 was measured for the bending radius from 8 to 11.5
mm at a wavelength of 1550 nm without connecting the polarization controller. The
bend loss for these bend radii with a length of 10 turns is presented in
Fig. 6(a) and the calculated results are presented as well
for a comparison. The experimental and theoretical results are in a satisfactory
agreement. Both results show the influence of the coating layer on the bend loss,
e.g., the bend loss does not decrease monotonically as the bending radius increases
as mentioned above.
Fig. 6. (a). Measured bend loss of SMF28 at wavelength 1550 nm with a length of 10
turns; (b) normalized polarization dependent loss.
To investigate the polarization dependence of the bending fiber under different
bending radii, the maximal and minimal values of measured bend loss are recorded
while tuning the polarization controller. It is not possible to find out the
respective bend loss for the TE and TM modes directly in this experiment with the
polarization controller, which instead offers the maximal and minimal bend loss
through adjusting the polarization controller. Therefore, the signs of the
theoretical normalized polarization dependent loss are used as the reference for the
experimental results. Corresponding results are presented in
Fig. 6(b). From the figure one can see that the theoretical
prediction have a good agreement with the measured normalized polarization dependent
loss.
Finally, we investigated the polarization dependence of the bend loss for the bending
fiber under different wavelengths. Corresponding to the simulation results presented
in
Fig. 3(b), the polarization dependent loss are measured for
(a) a fiber with a bending radius of 10 mm and a length of 20 turns and (b) a fiber
with a bending radius of 10.5 mm and a length of 10 turns. The measured polarization
dependence over the wavelength range from 1500 to 1600 nm are presented in
Fig. 7 with marks as squares and circles, respectively. For
convenience, the calculated results of the PDL are also presented in
Fig. 7 as the comparison. From
Fig. 7 one can see the measured PDL for bending radius of 10
mm is bigger than that for the bending radius of 10.5 mm as predicted by the
theoretical calculation. The calculated polarization dependent losses matched the
measured results. The theoretical predicted results for the bending radius of 10.5
mm have a better agreement with the measured PDLs as compared to the case of 10 mm.
The discrepancy between the calculated PDL and measured results could be caused by
the approximations made in the calculation. Calculation of bend loss and
polarization dependence for a singlemode fiber includes some approximations, e.g.,
the curved interface between the coating and cladding is treated as an infinite
plane and the light field within the core is approximated by the unperturbed field
of the straight fiber with infinite cladding as mentioned above. The calculated bend
loss and polarization dependence has an overall agreement with the experimental
results, i.e., they can not guarantee the accurate agreement for each bending radius
due to its limited accuracy. This can also be seen in
Fig. 6(b). For some bending radii, the calculated normalized
PDL has a good agreement with the experiment but for some bending radii there is a
discrepancy between the prediction and experiment. From
Fig. 6(b), one can also see that the bend radius of 10 mm
suffers a bigger discrepancy between the theoretical and experimental result as
compared to the case for bending radius of 10.5 mm.
Fig. 7. Measured polarization dependent losses for bending radius of 10 (20 turns)
and 10.5 mm (10 turns).