discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/8207842
CITATIONS READS
31 101
2 authors, including:
Svetislav Savovi
University of Kragujevac
84 PUBLICATIONS 520 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Svetislav Savovi on 09 March 2015.
Using the power-flow equation, we have examined the state of mode coupling in step-index plastic optical
fibers with different numerical apertures. Our results confirm that the coupling rates vary with the
coupling coefficient of the fibers as the dominant parameter, especially in the early stage of coupling near
the input fiber end. However, we show that the fibers numerical aperture has a significant influence on
later stages of this process. Consequently, equilibrium mode distribution and steady-state distribution
are achieved at overall fiber lengths that depend on both of these factors. As one of our examples
demonstrates, it is possible for the coupling length of a high-aperture fiber to be similar to that of a
low-aperture fiber despite the three-times-larger coupling coefficient of the former. 2004 Optical
Society of America
OCIS codes: 060.2310, 060.2400.
with signal modulation. This is but one example of
the need for an effective means of calculating the rate 0
P, z exp , (4)
of mode coupling, such as we present in this paper. 2 2
Use of geometric optics ray approximation to in-
vestigate mode coupling and to predict output-field
patterns has been reported.5 Employing the power- with 0 c, where 0 is the mean value of the
flow equation6 11 as well as the FokkerPlanck equa- incidence angle distribution, with a FWHM of
tion and the Langevin equation12 has yielded 22 ln 2 2.355 is the standard deviation.
predictions of these patterns as a function of launch This distribution is suitable for both LED and laser
conditions and fiber length. Coupling length Lc for beams.
achieving the EMD, the transition states along Lc, as
well as length zs for achieving the SSD are deter- 3. Numerical Solution
mined here by solution of the power-flow equation for In this paper we analyze mode coupling in three SI
SI POFs with various numerical apertures investi- POFs with different numerical apertures NAs used
gated earlier by Losada et al.3 in the experiment reported above.3 Each of the
three SI POFs had a 1-mm-diameter polymethyl-
2. Power-Flow Equation methacrylate core. The standard high-NA S-HNA
Gloges power-flow equation is6 fiber, HFBR-RUS500 from Hewlett-Packard, has a
NA of 0.47 with an attenuation of 0.22 dBm. PGU-
P, z D P, z CD1001-22E fiber from Toray has a similarly high
P, z , (1) NA, 0.5. It is an upper-grade high-NA U-HNA fi-
z
ber with a lower attenuation of 0.18 dBm. PMU-
where P, z is the angular power distribution; z is CD1002-22-E fiber, also from Toray and with an
the distance from the input end of the fiber; is the improved bandwidth, has a low NA LNA of 0.32 and
propagation angle with respect to the core axis; D is an attenuation of 0.20 dBm. In all cases the refrac-
the coupling coefficient, assumed constant7,8; and tive index of the core is n1 1.492. The number of
is the modal attenuation. The boundary conditions modes at 665 nm is3 N 22a2NA22 2.8,
are Pc, z 0, where c is the critical angle of the 2.5, 1.1 106 for the U-HNA, S-HNA, and LNA
fiber, and DP 0 at 0. Condition Pc, z fibers, respectively. These large numbers of modes
0 implies that modes with infinitely high loss do may be represented by a continuum, as required for
not carry power. Condition DP 0 at 0 application of Eq. 2. Coupling coefficient D is re-
indicates that the coupling is limited to the modes quired in that equation. We found that D 8.7,
propagating with 0. 3.3, 2.8 104 rad2m for the S-HNA, U-HNA, and
Except near cutoff, the attenuation remains uni- LNA fibers, respectively,3 values that we have
form at 0 throughout the region of guided adopted in this study.
modes, 0 c Ref. 8; it appears in the solution as Figures 13 illustrate our numerical solution of the
multiplication factor exp0 z, which also does not power-flow equation by showing the evolution of the
depend on . Therefore need not be accounted normalized output power distribution with fiber
for when one is solving Eq. 1 for mode coupling, and length. We show results for four input angles 0
that equation reduces to8,9 0, 5, 10, 15 for the S-HNA and U-HNA fibers and
for three input angles 0 0, 5, 10 for the LNA
P, z D P, z 2P, z fiber measured inside the fiber. Input angles near
D . (2) critical values of c 19.5, 18.4, and 12.4 for the
z 2
U-HNA, S-HNA, and LNA fibers, respectively, have
The solution of Eq. 2 for the steady-state power not been included because attenuation constant
distribution is given by8 cannot be neglected near the critical angles13 when
the original power-flow equation 1 has to be solved.
We selected a Gaussian launch-beam distribution
P, z J 0 2.405 exp 0 z, (3) with a FWHM of 2.5 by setting 1.06 in Eq. 4.
c
Fiber length z was incremented from zero until the
where J0 is a Bessel function of the first kind and zero SSD was achieved. A new curve or graph of the type
order and where 0 m1 2.4052Dc2 is the atten- shown in Fig. 1 was obtained for every increment in
uation coefficient. We used this solution to test our variable z. The peak of the intensity distribution
numerical results for a fiber length at which the decreased and its width increased with each step.
power distribution becomes independent of the To facilitate easier comparisons, we normalized every
launch conditions. peak before each new step. We used step lengths
To obtain a numerical solution of power-flow equa- 0.05 and z 0.0002 m to achieve stability of
tion 2 we used the explicit finite-difference method our finite-difference scheme.9,14
4. Results 1d, 2d, and 3d: zs 45, 50, 100 m for the
The radiation patterns for three different fibers are S-HNA, LNA, and U-HNA fibers, respectively.
shown in Figs. 13. It is not until the fiber coupling For the S-HNA, LNA, and U-HNA fibers, respec-
lengths Lc as shown in Figs. 1c, 2c, and 3c are tively, Figs. 1d, 2d, and 3d show normalized
achieved that all the mode distributions shift their curves of the output angular distribution obtained by
midpoints to 0 from the initial value of 0 at the the power-flow equation by the explicit finite-
input fiber end, producing the EMD at Lc 15, 18, difference method solid curves as well as the steady-
35 m for the S-HNA, LNA, and U-HNA fibers, respec- state analytical solution of Eq. 2 squares, where
tively. The coupling continues further along the fi- 0 0.01567, 0.00748, 0.00561 m1. The two solu-
ber beyond the Lc mark until all distributions widths tions are in good agreement, with the relative error
equalize and the SSD is reached at length zs in Figs. below 0.2%. As shown in our earlier research,9
these reasonably small relative errors for the largest
fiber lengths analyzed demonstrate further that the
explicit finite-difference method used is a sufficiently
accurate and effective method for solving the power-
flow equation.
Based on Figs. 13, the rate of mode coupling is
highest for the short S-HNA fiber in Fig. 1a as had
been determined previously in experiments with
power loss in 4-m long fibers3. This fiber also had
the highest coupling coefficient, D 8.7 104
rad2m. The LNA and U-HNA fibers exhibited vir-
tually identical but lower initial coupling rates be-
cause they have mutually similar but lower coupling
coefficients: D 2.8 104, 3.3 104 rad2m,
respectively. Whereas it was expected that the
higher coupling coefficient would lead to higher cou-
pling rates, such is not necessarily true for longer
fiber lengths, as we demonstrate next.
Despite the similar coupling coefficients of the LNA
and U-HNA fibers, these fibers have significantly dif-
Fig. 2. Normalized output angular power distribution at several
ferent coupling lengths Lc for the EMD and zs for
locations along the LNA fiber calculated for three Gaussian input the SSD: Lc 18 m and zs 50 m for LNA and
angles 0: 0 solid curves, 5 dashed curves, and 10 dotted Lc 35 m and zs 100 m for U-HNA Figs. 2c and
curves with a FWHM of 2.5 for z 4, 10, 18, 50 m squares 2d and Figs. 3c and 3d, respectively. The rea-
represent the analytical steady-state solution. son for this is that the fibers NAs differ substantially.