Design and Fabrication of All-Fiber Flat-Top Interleaver for DWDM Applications A dissertation submitted in partial fulfillment of the requirements for the degree of Master of Technology (Optoelectronics and Optical communication) by S Ravi Kumar (2005JOP2128) Under the guidance of Prof. B. P. Pal & Dr. R. K. Varshney Departments of Physics and Electrical Engineering Indian Institute of Technology Delhi May 2007
64
Embed
Design and Fabrication of All Fiber Flat Top Inter Leaver in DWDM Applications
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Design and Fabrication of All-Fiber Flat-Top Interleaver for DWDM Applications
A dissertation submitted in partial fulfillment of the requirements for the
degree of
Master of Technology
(Optoelectronics and Optical communication)
by
S Ravi Kumar
(2005JOP2128)
Under the guidance of
Prof. B. P. Pal & Dr. R. K. Varshney
Departments of Physics and Electrical Engineering
Indian Institute of Technology Delhi May 2007
i
CERTIFICATE
This is to certify that the dissertation entitled “Design and fabrication of all-fiber flat-top
interleaver for DWDM applications” being submitted by Mr. S Ravi Kumar
(2005JOP2128) to the Department of Physics and Department of Electrical Engineering,
Indian Institute of Technology, Delhi, in partial fulfillment of the requirement for the
degree of Master of Technology in Optoelectronics and Optical Communication, is a
record of bona fide work carried out by him under our supervision.
The results contained in this report have not been submitted elsewhere for any degree or
diploma.
Prof. B. P. Pal Professor Department of Physics IIT Delhi
Dr. R. K. Varshney Principal Scientific Officer Department of Physics IIT Delhi
ii
ACKNOWLEDGEMENTS
I am greatly indebted to my supervisors Professor B. P. Pal and Dr. R. K. Varshney for
their invaluable support, guidance and utmost interest during the course of this project. In
ways many more than I can express, they have motivated me to work hard and taught me
to think independently. I have truly learnt much from their optimistic way of thinking.
But for their helpful inputs, this project would have been far from completion. Thanks
also to Prof. M. R. Shenoy for his kind help and suggestions.
I must thank Dr. Naveen Kumar for the time and effort he has spared for me and the
valued inputs that were given by him. I express my gratitude to Mr. Nagaraju for being
with me in the lab and creating a work environment, all during my lab work. He has
taught me the coupler fabrication technique among other useful things. Thanks to Mr.
Raja Vinayagam for providing me with the lab equipment whenever I needed.
Special thanks to Prof. Vinod Chandra and Electrical Engineering Department for
allowing me to use lab equipment which was greatly useful in carrying out this project.
I would also like to thank my classmates Krishnan, Pavan, Kapil, Praveen, Saha, Anil,
Renuka, Abhinav, Sumitha, Deepak, Anoop, Sudhir and Vidya for their support
throughout my stay in IIT.
Last, but not the least, I thank my parents and other family members for their constant
support and influencing me in every way of life.
iii
ABSTRACT
The deployment of optical networks for communication has lead to rapid development in
the area of communication. This has given rise to newer applications with greater demand
for data bandwidth. Dense wavelength division multiplexed (DWDM) optical networks
with high spectral efficiency are indispensable means to cater to present day need of
bandwidth. In this work, the focus has been on design and fabrication of all-fiber flat-top
interleaver for DWDM applications.
The concept of DWDM is introduced in the first chapter of this thesis. A fair idea of
DWDM networks is essential to understand the role of an interleaver in DWDM
networks. To enhance research and development in this field, technical specifications
have been recommended. These have been briefly discussed herein.
We have studied and analyzed various design approaches used for filters in general, and
interleavers in specific. Simulation was carried out to calculate structure parameters using
these approaches. The results have been verified by comparing with data from literature.
Design to compensate dispersion has also been studied.
A two-stage flat-top interleaver was fabricated in the lab and was characterized. The
technique to implement the interleaver has been discussed.
Polymer based optical components have several advantages and therefore have been
receiving much attention in recent years. An interleaver based on Y-junctions instead of
directional couplers, using polymer technology has been reported. But it does not have
appreciable characteristics of an ideal interleaver, like flat passband and high extinction
ratio. We have proposed and worked out a design to achieve these characteristics in a
polymer Y-junction based interleaver. This structure has been analyzed and the structure
parameters that would give the optimum characteristics have been derived.
2.5.1 Delay line in the upper arm of the second stage ....................................... 12 2.5.2 Delay line in the lower arm of the second stage ....................................... 14
Table 7: Effect of additional stages in all-pass filter for equalizing dispersion................ 30
Table 8: Typical measured characteristics of the couplers fabricated .............................. 34
Table 9: Optimized splitting ratios of the 3 couplers K1, K2 and K3 in a 2-stage MZI-
based interleaver when the delay line effnLL 2/2 12 λ±Δ=Δ in the upper arm ....... 53
Table 10: Optimized splitting ratios of the 3 couplers K1, K2 and K3 in a 2-stage MZI-
based interleaver when the delay line 12 2 LL Δ=Δ is in the lower arm.................... 53
Table 11: Optimized parameters for a 2-stage MZI configuration using Y -junctions when
the delay is in the upper arm..................................................................................... 54
Table 12: Optimized parameters for a 2-stage MZI configuration using Y -junctions when the delay line is in the lower arm line is in the lower arm……………………………….54
i
1
Chapter 1
Introduction
To meet bandwidth demand and make optimum use of existing amplifier
bandwidth, dense wavelength division multiplexing (DWDM) systems must offer higher
channel counts at narrower channel spacing and the requirements are on an incessant
increase [1]. Use of 160 channels with 40 Gbps bit rate per channel is quite common
today. A number of different DWDM technologies exist to meet the needs of system
designers, but each forces design tradeoffs in terms of narrowness of channel spacing,
cost, reliability, and manufacturability. Nowadays, interleaver technology is being
deployed as it allows designers to achieve narrow channel spacing with mature
technology.
1.1 DWDM The predominant aspect of a DWDM communication link is that several
wavelengths (at least four), each carrying digital signals @ ≥2.5 Gbit/s, could be sent
through a single-mode fiber simultaneously within the 1530-to-1610-nm gain bandwidth
of a fiber amplifier with inter-channel spacing ≤ 200 GHz. The key features of DWDM
are
Transmission capacity upgradation: As demand for bandwidth grows, each wavelength
channel can carry data at independent data rates and can be upgraded independently.
Transparency: Each optical channel can carry data in any transmission format and at any
data rate. This makes the network design and engineering simpler and more flexible as
there is no need for common signal structure.
Wavelength routing and switching: Wavelength can be exploited as a new dimension for
routing besides space and time. Further, wavelength-switched architectures allow
2
reconfiguration of optical layer. Optical add/drop multiplexers, optical cross connects and
wavelength converters are used for implementation of these networks.
1.2 DWDM Operation DWDM networks are sought after to utilize the enormous transmission bandwidth
of an optical fiber beyond what could be achieved by simply increasing the bit rate of a
transmitter-receiver pair. In DWDM, multiple transmitters, each at a different
wavelength, are combined onto a single fiber by a multiplexer (Mux). At the other end, a
demultiplexer (DeMux) separates out the wavelengths into separate receivers. In this
way, multiple transmitter-receiver pairs share the same fiber. The schematic of a typical
modern DWDM optical link is shown in Fig1.
Figure 1: Schematic of a typical DWDM optical fiber long-haul link[11]
A genre of high-quality semiconductor lasers, called distributed-feedback (DFB)
lasers, is typically used in a DWDM system. Electronic digital inputs modulate individual
lasers, each emitting at a different peak wavelength. The spectral linewidth of modulated
output of a DFB laser is typically of the order of 10-3 nm. An array of DFB lasers, each
tuned to a different but well-defined and standardized wavelength are used. Interference
among the wavelength channels is avoided and the integrities of the independent signals
Laser Diode at ITU Wavelengths
OE
Multiplexer
EDFA EDF
Long Bragg Circulator
EDF
Demultiplexer OE
OE
OE
OE
OE
λ
λ
λ
λ
λλ
λ3
Add-Drop multiplexer
Optoelectronic Converter
Isolator
λ3
3
from each source are maintained for subsequent filtering and conversion to electrical
signals at the receiving end by sufficiently spacing them from their neighbours.
1.3 Technical Requirements and Specifications To coordinate and facilitate the growth of DWDM fiber links, International
Telecommunication Union (ITU) has set certain wavelength standards, referred to as ITU
wavelength grids, for DWDM optical transmission systems. As per ITU standards (ITU-
T recommendation G.692), the reference frequency ( 0ν ) is chosen to be that
corresponding to the Krypton line i.e. 193.1 THz. (≡ 1552.52 nm in wavelength), and the
chosen channel spacing are expected to follow the relation
νΔν ±0 (THz) = 193.1 ± 0.1I, where I is an integer.
The recommended standard channel spacings are 200 GHz (≡ 1.6 nm), 100 GHz (≡ 0.8
nm), 50 GHz (≡ 0.4 nm), and 25 GHz (≡ 0.2 nm). It is to be noted that ITU standards
specify fixed frequency spacing, rather than constant wavelength spacing. Variations in a
laser’s peak wavelength lead to crosstalk between adjacent channels. Therefore it is
necessary to keep the variations in wavelength small compared to the spacing between
adjacent channels i.e. a laser signal should be prevented from “wandering” into an
adjacent channel in a DWDM stream. One would have to use the manufacturer’s design
specifications for a particular DWDM system. Those specifications may be tighter than
the 10% allowed by ITU.
1.4 Role of Interleavers/De-interleavers in Present day
DWDM Links Due to the difficulties encountered in attaining high-speed electronic components at
speeds beyond 40 Gbit/s, increase in the number of channels has evolved as the best near-
term option to achieve high spectral efficiency.. This can be done in two ways
1. Increasing the operational communication bandwidth per channel.
2. By decreasing the inter-channel spacing.
4
These two aspects are combined to define spectral efficiency, which refers to the
amount of information that can be transmitted over a given bandwidth.
A DWDM optical network requires a variety of passive and active components like
wavelength multiplexers/de-multiplexers, EDFAs, gain flattening filters for EDFAs, add-
drop multiplexers, dispersion compensators, bandpass filters, lasers, and so on. Since,
most of the current DWDM networks operate in the C-band, which corresponds
approximately to the gain band from 1530 nm to 1565 nm of an erbium doped fiber
amplifier (EDFA), increasing the operational communication bandwidth involves
replacing the present components. The newer components would be costlier as they
would require technologies different from the ones used presently. Also, additional
amplifiers have to be used for the other bands namely, L-band (1570 nm - 1620 nm) and
S-band (1480 nm ∼ 1525 nm).
The alternative option, i.e. to decrease the inter-channel spacing, is generally preferred.
But, this would still mean replacing some components for e.g., filters with narrower
passbands would be required at the receivers, which is also expensive. Interleavers/de-
interleavers provide an easy solution in this regard and therefore have received much
attention.
5
Chapter 2
Interleaver Types and Design Methods 2.1 Interleaver Interleaver/de-interleaver is an optical filter with a periodic response that is capable of
separating a comb of densely spaced signal channels into several sets of channels with
wide channel spacings or doing the converse [3]. Because of the periodic nature of
interleave filter, its transfer function consists of fewer number of Fourier components
compared to, say, add/drop filters. For this reason, the interleaver is simpler to realize
than other filters.
2.2 Interleaver Types As a functional block, interleavers can be considered to be of different types [4]
1:2 interleaver: It combines two or more input streams of wavelength-channels
having a constant spacing νΔ in the frequency domain into a single dense stream of
channels with separation 2/νΔ at the output (see Fig. 2a ). When operated in the reverse
direction, it separates the even and odd signal channel streams.
1:4 interleaver: The natural extension of the 1:2 interleave filter is the periodic
separation of one in every 2n channels, such as 1: 4 demultiplexer depicted in Fig. 2b.
This device has four different output ports. This kind of interleaving (i.e. 1:2n) can also be
achieved by cascading 2n-1 stages of 1:2 interleave filters.
Banded interleaver: This type of interleaver separates/combines, one stream of
wavelength channels into two bands of channels, coming from the two different output
ports, periodically. It is difficult to fabricate such device in practice because the
6
requirements to achieve sharp skirt for the filter to meet the ITU specifications are very
tight.
Asymmetric interleaver: The other variation in contrast to the three types of
interleaving mentioned above is the asymmetric interleaver which separates one
wavelength from the stream of wavelength channels such that one channel appear at one
output port whereas the remaining (n-1) channels exit from the other output port
periodically i.e. it separates one channel out of n.
Figure 2: Schematic of different interleavers. a) 1:2 interleaver b) 1:4 interleaver c) Banded interleaver d) Asymmetric type interleaver
2.3 Single-stage MZI Interleaver An all-fiber single-stage unbalanced MZI can be formed by splicing two 2×2 fiber
couplers in such a way that the lengths of the two arms are slightly unequal [5] by an
amount 1LΔ . The differential path length ( 1LΔ ) is equivalent to a differential delay line in
one of the arms. The corresponding differential phase delay 1ϕ suffered by the channel
wavelength λ is given by
λΔπϕ /2 11 Lneff=
1:2
Banded 4:8
a)
c)
d)
λe
λb2
Asym 1:8
λa1
λa7
λo
1:4 b)
λa
λb
λc
λd
λb1
λ……..
λ……..
λ……..
λ……..
(2.1)
7
where )(λeffn is the effective index of the guided mode at λ .
Figure 3: Schematic of an all-fiber MZI based interleaver, realized through concatenation of two 3-dB couplers; ΔL1 represents a delay line in the upper arm of the interferometer.
The transmission characteristics of an interleaver based on MZI can easily be described
in terms of transfer matrices of the individual couplers and the differential delay lines [6].
From the coupled mode equations for the electric field amplitudes, the transfer matrix for
the coupler is given by
⎟⎟⎠
⎞⎜⎜⎝
⎛−
−=
cjsjsc
M coupler
where )cos( zc κ= , and )sin( zs κ= ;κ is the coupling coefficient and z is the interaction
length of the couplers. The transfer matrix corresponding to the differential phase delay
1ϕ between the two arms is given by
⎟⎟⎠
⎞⎜⎜⎝
⎛=
1001
1
ϕ
Δ
j
Le
M
Thus, product of the transfer matrices for the couplers with those of the differential delay
line would yield the transfer function of an unbalanced MZI (shown in Fig. 7), and is
given as
12 1 couplerLcouplerMZI MMMM Δ=
If 1E and 2E are input fields to the MZI at Ports 1 and 2, respectively, then the output
fields TE and CE at Ports 3 and 4 can be expressed as
Port 1
Port 2
3 dB 3 dB
Port 4
Port 3Coupler 1 Coupler 2
)( 22 λP
LΔ
)( 11 λP
4P
),( 213 λλP
(2.4)
(2.2)
(2.3)
8
⎟⎟⎠
⎞⎜⎜⎝
⎛=⎟⎟
⎠
⎞⎜⎜⎝
⎛
2
1
EE
MEE
MZIC
T
⎟⎟⎠
⎞⎜⎜⎝
⎛⎟⎟⎠
⎞⎜⎜⎝
⎛−
−⎟⎟⎠
⎞⎜⎜⎝
⎛⎟⎟⎠
⎞⎜⎜⎝
⎛−
−=⎟⎟
⎠
⎞⎜⎜⎝
⎛
2
1
11
11
22
22
1001
EE
cjsjsce
cjsjsc
EE j
C
Tϕ
where ii zc )cos(κ= , iz)sin(si κ= , i = 1, 2.
Thus, if *1 1 1 1 1 1( ) ( ) ( )P E Eλ λ λ= =1 and )()()( 2
*22222 λλλ EEP = =1 represent the normalized
powers at the two input ports, corresponding to the channel wavelengths 1λ and 2λ ,
respectively, then the fractional output powers at the throughput and coupled ports are
respectively given by
* 2 21 1 1 21 1 2 2
( ) ( )( )sin ( )cos2 2T T TP E E P Pϕ λ ϕ λλ λ= = +
* 2 21 1 1 21 1 2 2
( ) ( )( ) cos ( )sin2 2C C CP E E P Pϕ λ ϕ λλ λ= = +
The splitting ratios of the couplers 1 and 2 in the MZI configuration have been assumed
to be 50:50. In order to achieve wavelength interleaving by an unbalanced MZI
configuration, the differential phase 1ϕ has to satisfy either of the following conditions:
πλϕ )12()( 11 += n and πλϕ n2)( 21 = ; (PC = 0, PT = Pmax)
or
πλϕ n2)( 11 = and πλϕ )12()( 21 += n ; (PT = 0, PC = Pmax)
where n is an integer. From Eq. (2.9) and Eq. (2.10), we can observe that
πλϕλϕ =− )( ( 21 11 ) .
Thus, by substituting the values of )( 11 λϕ and )( 21 λϕ in Eq. (2.1), we get
λΔλλ
Δeffn
L2
211 =
Physically, if the DWDM signal channels are input at Port 1, the signal wavelengths that
suffer a differential phase delay of π)12( +n will exit through Port 3, and the
wavelengths that suffer a differential phase delay of πn2 will exit from Port 4. These
wavelengths correspond to the peaks in the spectral response. The single-stage MZI
exhibits a sinusoidal spectral response at the output ports 3 and 4 that are complementary.
(2.5)
(2.7)
(2.6)
(2.8)
(2.10)
(2.9)
(2.11)
(2.12)
9
The wavelength separation between the consecutive peaks at any output port represents
the free spectral range (FSR) of the MZI; the FSR is determined solely by 1ϕ between the
propagating signals in the two arms of the MZI. If we change 1LΔ by an amount
effn2/λ (which introduces an additional phase change ofπ ), the spectral response at the
output ports gets reversed.
The spectral response of the single-stage MZI is shown in Fig.7 of next chapter.
2.4 Desired Characteristics of an Interleaver
Low insertion loss: Any component used in a DWDM network has to be of low insertion
loss since a lossy device would lead to attenuation of signal and thus degradation of
signal-to-noise ratio, even after amplifying. Any initial imbalance in channel-power will
have a cascading effect, as the signal is normally required to pass through multiple stages,
and may affect the signal-to-noise performance of the system
Wide Pass band: The wavelength emitted by the laser source in the transmitter can vary
slightly due to changes in ambient conditions like temperature. The components used
should allow for this small variation.
Wide Rejection band: The power in the rejection band should be considered as noise to
the channel from other channels (crosstalk). Also this would reduce the power in the
passband leading to higher insertion loss.
High extinction in the Rejection band: The response at the rejection band is the power
that would be coupled as crosstalk from this channel to other channels. Hence, the
attenuation in the rejection band should be as high as possible.
Good wavelength accuracy: The interleaver’s spectral response should not change over
time and should be fixed to the DWDM wavelength grid specified by the ITU. The
variation in spectral response can lead to severe performance degradation through
attenuation of signal as well as crosstalk from adjacent channels.
10
Low dispersion: The filters and other components used in DWDM networks have strict
requirement of low dispersion since very high bit rates (40 Gbps, typically) are used.
Low Polarization dependency: The various kinds of polarization dependencies like
polarization mode dispersion (PMD), polarization dependent loss (PDL), and polarization
dependent wavelength shift (PDλ) cause loss of signal and have to be avoided.
Uniformity across channels: The interleaver should have uniform spectral response over
the optical bandwidth used for DWDM, since the transmission characteristics should be
independent of the wavelength used for transmission.
2.5 Two-stage MZI Interleaver
A 2-stage MZI configuration has been proposed to meet the just discussed requirements
[7]-[9]. It can be realized by concatenating three couplers through two differential delay
lines in the two stages, as shown in Fig. 4. In order to achieve a uniform flattop response
with minimum insertion loss, one has to choose optimum splitting ratios for the couplers
K1, K2, and K3, as well as magnitudes of the differential delay lines 1LΔ and 2LΔ .
Figure 4: A 2-stage MZI configuration realized by concatenating three couplers K1, K2, and K3; the two delay lines 1LΔ and 2LΔ are introduced in the upper arms of the first and second stages, respectively
The phase difference between the fields E13 and E14 at ports 3 and 4 of the single-stage
MZI configuration (see Fig. 3) plays an important role in deciding the magnitude of the
delay line to be introduced in the second stage for achieving a flattop spectral response of
)( 11 λP
4P3 dB
K1 K2
)( 22 λP
1LΔ
K3
2LΔ),( 213 λλP
11
the 2-stage MZI; Considering unit input at only port1, the resultant output fields at Ports
3 and 4 are given as [5, 6]
212113 1 sseccEE jT −== ϕ
)( 212114 1 csescjEE jC +== ϕ
If the splitting ratio of the first coupler is 50:50, and that of second coupler is more than
50: 50 (i.e. s2 > c2), the phase difference (θ) between the output fields E13 and E14 is given
by [11]
1413
14131 .cos
EEEE
−=θ (s2 > c2)
Likewise, if the splitting ratio of the first coupler is 50:50 and that of second coupler is
less than 50:50 (i.e. c2 > s2), the phase difference (θ) between the output fields E13 and E14
is given by [11]
1413
14131 .cos2
EEEE
−−= πθ (c2 > s2)
It is imperative from the Eq. (2.15) and (2.16) equations that the magnitude of the
optimum delay-line in the second stage would depend on whether the splitting ratio of the
second coupler is more, or less than 50:50. Often a compromise is made between
achieving a flattop passband and a low insertion loss, since additional filtering elements
are usually added to an original sinusoidal passband shape to achieve a uniform flattop
response.
Position of the Second Delay Line
To realize a 2-stage MZI configuration with a flattop response, a second 2×2 coupler is
concatenated to a single-stage MZI with an optimum differential delay line in the second
stage. There are two possible choices for a designer- the second delay line could be
introduced either in the upper arm or in the lower arm of the second stage of the 2-stage
MZI [12, 10]. Though these two optional configurations appear to be apparently
identical, the performance of these two configurations are dictated by the splitting ratios
of the constituent couplers as well as the phase differences between the fields at ports 3
and 4, and also by the magnitude of the delay lines in the second stage.
(2.13)
(2.14)
(2.15)
(2.16)
12
2.5.1 Delay line in the upper arm of the second stage The transfer matrix corresponding to this 2-stage MZI configuration is given by [9, 12]
Figure 9: Simulated spectral response in linear scale at Port 3 of the MZI-based interleaver at 1550 nm.
ΔL = 2.1 mm corresponding to channel spacing of 100 GHz
Figure 10: Simulated spectral response in log scale at Port 3 of the MZI-based interleaver at 1550 nm.
ΔL = 2.1 mm corresponding to channel spacing of 100 GHz
22
The spectral response of the single stage MZI interleaver discussed herein has been
simulated in Matlab using the transfer matrix approach for variable differential length
corresponding to variable channel spacing.
3.1.2 Two-stage MZI Interleaver Spectral Response The spectral response of the 2-stage MZI interleaver has been simulated in Matlab using
the transfer matrix approach. The simulation was carried out by varying the differential
lengths as well as the coupling ratios of the couplers.
The following observations were made:
1. The transfer function corresponding to an interleaver was obtained only for
certain values of differential lengths.
2. The periodicity of the interleaver (channel spacing) depends on the differential
length.
3. The response when the second differential length is placed on the same side as the
first was different from the response obtained when it is placed on the opposite
side with respect to the first delay line.
4. Maximum throughput power at the centre of the passband was obtained only
when the first coupler has 50:50 coupling ratio.
1.549 1.55 1.551
0.2
0.4
0.6
0.8
Wavelength (µm)
Nor
mal
ized
Tra
nsm
ittan
ce
1.549 1.55 1.551-35
-30
-25
-20
-15
-10
-5
Wavelength (µm)
Nor
mal
ized
Tra
nsm
ittan
ce (d
B)
Figure 11: Simulated spectral response of a 2-stage MZI interleaver for the following parameters:
Coupler1- 50:50, Coupler2- 68:32, Coupler3- 4:96 at 1550 nm. ΔL2 = 2ΔL1; ΔL1 = 2.1mm for a channel spacing of 100GHz
23
3.1.3 Design of Interleaver Based on the Passband Ripple Following S.W.Kok, et al, [10] simulation has been done for the design of flat-top interleaver based on the passband ripple. The results are summarized in the following graphs.
0 0.02 0.04 0.06 0.08 0.1 0.12-5
0
5
10
15
20
25
30
Passband Ripple (dB)
Pas
sban
d w
idth
at 0
.1 d
B (
GH
z)
Figure 12: Variation of 0.1dB-passband width with increasing passband ripple for interleaver at 1.55μm
and 50 GHz channel spacing
0 0.1 0.2 0.3 0.4 0.529
30
31
32
33
34
35
36
Passband Ripple (dB)
Pas
sban
d w
idth
at 0
.5 d
B (G
Hz)
Figure 13: Variation of 0.5dB-passband width with increasing passband ripple for interleaver at 1.55μm
and 50 GHz channel spacing
24
0 0.005 0.01 0.015 0.020
2
4
6
8
10
12
14
16
Passband Ripple (dB)
Sto
pban
d w
idth
at 2
5 dB
(GH
z)
Figure 14: Variation of 25 dB-stopband width with passband ripple for interleaver at 1.55 μm and 100 GHz channel spacing
0 0.1 0.2 0.3 0.4 0.5-150
-100
-50
0
Passband Ripple (dB)
Cro
ss T
alk
leve
l (dB
)
Figure 15: Variation of crosstalk with passband ripple for interleaver at 1.55 μm and 100 GHz channel spacing
From the above plots (Figs. 12-15) the following conclusions can be drawn:
1. Passband width increases with passband ripple. This indicates that there is a trade-
off between the insertion loss and the passband width.
25
2. Stopband width also increases with passband ripple. This indicates that there is a
trade-off between the insertion loss and the stopband width as well.
3. However, the crosstalk also increases with passband ripple.
Also, for easy visualization of the effect as well as for better insight, a graphic user-interface has been developed using Matlab.
Figure 16: Matlab Graphics User Interface to calculate the structure parameters and the interleaver characteristics based on channel spacing and the specified passband ripple
26
3.1.4 Synthesis of Two-port Lattice Form Optical Delay Line
Circuits
Using the synthesis method proposed by K. Jinguji, et al [13], filters of various kinds can
be designed for varied purposes. Simulation has been done to calculate the structure
parameters for these designs. To validate the results, the simulation has been done for the
filters proposed therein and the structure parameters have been found to match very
closely with the results in literature.
-40 -20 0 20 40-50
-45
-40
-35
-30
-25
-20
-15
-10
-5
0
Tran
smitt
ance
(dB
)
Relative optical frequency (GHz)
Figure 17: Linear Chebyshev filter designed using optical delay line circuits for channel spacing of 100 GHz at 1.55 μm
Table 1: Simulated structure parameters for the designed Linear Chebyshev filter
Figure 19: Transmittance and relative group delay in a group-delay dispersion equalizer designed using optical delay line circuits for a channel spacing of 12.5 GHz at 1.55 μm
For optimum coupling ratiosFor a change of 5% in 1st coupling ratioFor a change of 5% in 2nd coupling ratioFor a change of 5% in 3rd coupling ratio
32
1550 1550.2 1550.4 1550.6 1550.8 1551-40
-35
-30
-25
-20
-15
-10
-5
0
Wavelength (nm)
Tran
smitt
ance
(dB
)
For optimum delay lengthsFor a change of 1µm in 1st delay lengthFor a change of 1µm in 2nd delay length
Figure 22: Spectral response when deviation of delay lengths is 1µm from the optimum value of 2.1mm
The simulation results showed that the tolerance towards delay lengths was minimal and
very accurate control over the lengths is required. However, there was not appreciable
deviation from optimum response due to variation in coupling ratios within the
achievable precision.
33
3.2 Experimental Work 3.2.1 Fabrication and Characterization of Fused Fiber Couplers To construct a 2-stage MZI interleaver, we need to fabricate couplers. Fused fiber
couplers were fabricated using the in-house made rig [21] (see Fig. 23). First, the
couplers were fabricated at 632.8 nm so as to learn the process of fabrication, since it is
easier to setup the experiment and also control the process at visible wavelengths.
Figure 23: Photograph of the fusion fiber coupler fabrication rig
Then, couplers at 1.55 μm were fabricated since these are required for the construction of
the MZI interleaver for DWDM applications.
The fabrication process involves stretching a pair of single-mode fibers together, which
are held in intimate contact with each other across a short unjacketed length, in a high
temperature (>1100 0C) oxy-butane flame. The fusion followed by stretching leads to
narrowing of the two fibers into a single biconical tapered junction. The tapered region is
essentially a multimode near-rectangular region with the core formed from the cladding
of the original fibers and the surrounding air acting as the cladding. Light from the input
fiber excites the two lowest order modes of this waveguide. Coupling of light from one
34
fiber to another is due to beating between the supermodes. This fabrication technique
consists of fusion, pulling and simultaneous tapering. Hence, it is also called as fuse-pull-
taper method in the literature. Due to symmetry in the structure formed, either end can be
used as input end. Hence these are known as bi-directional couplers. First appropriate,
stable and uniform flame is created. The various valves at different levels of the gas
ensure safety while providing fine pressure control. The temperature profile of the flame
affects the characteristics of the coupler significantly. A short length of fibers is
unjacketed using dichloromethane to soften the jacket. The unjacketed region is then
cleaned with acetone or iso-propyl alcohol. The fibers are twisted over each other in this
region and placed on the holder. Light is input at one end and the light from the two
output ports are monitored. The flame is positioned under the unjacketed region. The
fibers are then pulled slowly. When the required distribution of light between the output
ports is observed, the flame is withdrawn and the pulling is stopped a few seconds later.
Then the coupler is appropriately packaged. Table 8 depicts the typical measured
characteristics of the couplers fabricated in our lab.
Wavelength (nm) Coupling Ratio Excess Loss (dB) Directivity (dB)
1 632.8 48:52 1.32 33.4
2 632.8 46:54 1.02 44.2
3 632.8 48:52 1.17 40.5
4 1550 40:60 0.74 38.1
5 1550 95:5 1.10 32.2
6 1550 45:55 1.46 35.3
Table 8: Typical measured characteristics of the couplers fabricated using the in-house rig
35
But, as can be seen from the Table 8, the fabricated couplers at 1.55 μm have a loss much
greater than the expected loss (0.1 dB). The reasons for the loss could be attributed to the
following:
1. Only one detector was used to monitor power at two output ports, requiring the
fused fiber to be taken in and out of the flame a number of times leading to lossy
couplers.
2. It is difficult to manually control the pulling process. Due to malfunctioning of the
system used for controlling the pulling process through software, the fabrication is
now being done manually.
3.2.2 Techniques to Achieve Optical Path Difference between MZI
Arms As was seen from the simulation, to implement flat-top interleaver, precise control over
the phase difference between the arms of the MZI (which is caused due to delay lengths,
in this case) is required. There are two ways to control the phase difference i.e., by
modifying the mode effective index and changing the delay length. It requires a rather
complicated technique to induce a change in refractive index by ion bombardment, and
also, the induced changes in the refractive index slowly drift with time leading to drift in
the channel wavelength. Yonglin et al. [6] have reported technique to achieve phase
change by mechanical stretching of the fiber in one of the arms of the MZI, in real-time
while monitoring the wavelength response of the interleaver. But since these techniques
do not allow precise control on the length over which the fiber is stretched, they are at
best suited for coarse-WDM systems. Kumar et al. [11] have reported a technique of
selective heating and controlled stretching of fiber for tuning while real-time monitoring
the spectral response of the interleaver. We have adopted this method as it is convenient
and the response has been reported to be stable.
3.2.3 Tuning Channel Spacing and Channel Wavelengths In this section we derive relations necessary to estimate the change in the differential
path-length for tuning the channel wavelengths and inter-channel spacing. In order to
36
interleave two wavelengths 1λ and 2λ in an unbalanced MZI, the differential path length
1LΔ should satisfy
where 21 λλλΔ −= is the channel spacing. Suppose the differential path length is
changed 1LΔ to '1LΔ , and the corresponding new wavelengths, say, are 3λ and 4λ , then
' 3 41
eff2L
nλ λ
λΔ =
′Δ
where 43 λλλΔ −=′ .
From Eq. (3.1) and Eq. (3.2), we get the change in the differential path-length
' 1 2 3 41 1
eff2L L L
nλ λ λ λ λ λδ
λ λ′Δ − Δ
= Δ −Δ =′Δ Δ
ITU has fixed certain fixed channel wavelengths and inter channel spacing for DWDM
networks, to set standards for telecommunication networks spread across the world, To
conform to ITU recommendations, the channel wavelengths have to be tuned to match
the ITU defined grid, and, the channel spacing has to be tuned to the prescribed values
(viz. νΔ = 100 GHz, 50 GHz, or 25 GHz).
To tune the center wavelengths of the channels from 1λ and 2λ to 3λ and 4λ ,
respectively, without changing the channel spacing, we write
whereδλ is the shift in the wavelength channels, and 2/12121 )(2/)( λλλλλ ≈+= . Note
that for shifting the spectrum towards higher wavelengths
(i.e. δλλλδλλλ +=+= 2413 , ), 1'1 LL ΔΔ > , which implies that the longer arm of the
interferometer has to be further elongated by an amount Lδ . Similarly, to shift the
spectrum towards lower wavelengths (i.e. δλλλδλλλ −=−= 2413 , ), 1'1 LL ΔΔ < , and
hence the shorter arm of the interferometer is to be elongated by an amount Lδ . Eq. (3.5)
1 21
eff2L
nλ λ
λΔ =
Δ
(3.5)
(3.4)
(3.3)
(3.2)
(3.1)
37
shows that in order to tune an interleaver in accordance with the ITU wavelength grid for
DWDM systems (say, 100GHz spacing), a precise control over Lδ ∼ 1 µm is necessary.
To illustrate tuning of the channel spacing, consider a situation in which the channel
spacing λΔ has to be tuned to a new value λΔ ′ . Using Eqs. (3.1) and (3.2),
λΔλΔ
λΔΔδ′
−≈−=′11
2
2'11
effnLLL
where we have assumed 2/143
2/121 )()( λλλλλ ≈≈ . Thus, in order to increase the channel
spacing, one has to elongate the shorter arm of the MZI by an amount L′δ . Similarly, to
decrease the channel spacing, one has to further elongate the longer arm of the MZI by an
amount L′δ . In this analysis, the higher order terms in λΔδλ, and λΔδλ , can be neglected
as the effects of these terms are insignificant.
3.2.4 Estimation of the Differential Delays in a Two-stage MZI If the splitting ratio of the second coupler is less than 50:50, then the differential delay
has to be introduced in the lower arm of the second stage so as to achieve the flattop
response. If the splitting ratio of the second coupler is greater than 50:50, then the
differential delay line has to be introduced in the upper arm of the second stage for
achieving flattop spectral response. In practice, when we fabricate an all-fiber interleaver,
there will be an initial differential delay between the upper and lower arms of the first-
and second stages. Before tuning, we must have a priori knowledge of the existing
differential delay between the arms of the interferometer and its location. Therefore, it is
important to identify which arm is longer, and by how much. In an all-fiber MZI (as
shown in Fig.4), the first delay line is used to achieve channel spacing, and it solely
determines the FSR of the configuration. The second delay line is used for obtaining the
flattop response. When we concatenate two fiber directional couplers to realize an MZI,
the FSR of this interleaver can be measured from the spectral response on an OSA, which
gives us the information about the magnitude of the first delay line 1LΔ . In order to
identify whether the upper arm or the lower arm of the first stage is longer, one may
elongate the upper arm by a small amount. This would result in either decrease or
(3.6)
38
increase in the channel, implying that the lower arm is either longer or shorter,
respectively. For example, if the upper arm was longer, then elongation of the upper arm
would result in a decrease in the FSR. But if the upper arm was shorter, then elongation
of the upper arm would result in an increase in the FSR. We then concatenate another
coupler to the single-stage configuration to realize the targeted two-stage MZI-based
interleaver.
To estimate the initial differential path length in the second stage (see Fig 4) and its
location, light is launched from port 5, and the output is measured at ports 1 and 2. This
configuration for couplers with optimum splitting ratios is then simulated for the two
cases when the second differential path length is located in the upper arm and in the
lower arm. By matching these simulated curves with the observed ones on the OSA, we
can know whether the upper or the lower arm of the second stage is longer. In this way,
one can get the information required to realize a flattop interleaver before tuning.
3.2.5 Implementation of 2-stage MZI Interleaver The first part of the experiment consists of constructing the first MZI stage. For this, two
fused fiber couplers were used which were fabricated using our in-house developed
coupler-fabrication rig. This forms a single-stage all-fiber MZI (Fig 3).
Figure 24: Experimental set-up used to measure the wavelength response of an all-fiber MZI
Port3Optical fiber
Port1 Coupler 1 Coupler 2
ΔL1
Port4
OSA Broadband EDF Source
39
1520 1530 1540 1550 1560 1570-40
-35
-30
-25
-20
-15
Pow
er (d
Bm
)
Wavelength (nm)
EDF source full spectrum
Figure 25: Photograph of the two-stage MZI interleaver that was fabricated in the lab
A broadband super-fluorescent EDF source was used as the source at the input (Port 1) of
the MZI. The source was found to have a nearly flat spectral response in the range 1548-
1558 nm.
Figure 26: Power spectrum of EDF broadband source used for the experiment
40
1549.0 1549.5 1550.0 1550.5 1551.0
-39.75
-39.70
-39.65
-39.60
-39.55
-39.50
-39.45
-39.40
Pow
er (d
Bm
)
Wavelength (nm)
EDF source over 2nm range
The fibers at the output ports of the MZI were connected to an OSA using fiber
connectors. Figure 27 shows the spectral response of the unbalanced MZI. In order to
achieve tuning of the FSR, the required differential phase delay is introduced by selective
heating and controlled stretching of the longer arm of the MZI; a small unjacketed
segment of the fiber was exposed to an oxy-butane micro-flame, while simultaneously
monitoring the spectral response on the OSA. When the FSR was close to but still larger
than the required channel spacing, the flame burner was withdrawn and the elongation
process was stopped after a few seconds. This was done to avoid any sagging of the
fiber. However since the spectral response was not stable when the fiber was above the
flame, the fiber was withdrawn several times and the FSR was measured when the arms
of the MZI were away from the flame to ensure that the spectral response was stable. By
slowing down the pulling rate, the interleaver was also tuned to standard ITU wavelength
grid. The wavelength response of the fabricated MZI was then recorded in the OSA (see
Fig. 29).
Figure 27: Power spectrum of EDF broadband source over a range of 2 nm
To achieve tuning of wavelength response, the heating and elongation was essentially
done over a fiber length of about 2cm, along which the variation in the overall diameter
of the fiber (125 μm) was ∼ 15 μm. For this extent of tapering, the physical parameters of
41
1549.32 1550.12 1550.92-54
-52
-50
-48
-46
-44
-42
Pow
er (d
Bm
)
Wavelength (nm)
1541.34 1542.14 1542.94 1543.74-49
-48
-47
-46
-45
-44
-43
-42
Pow
er (d
Bm)
Wavelength (nm)
the fiber do not alter appreciably, and the fiber still retains its robustness. The device was
eventually lifted off the tuning rig and the arms of the MZI were appropriately protected
and heat-shielded to circumvent temperature-induced fluctuations in the wavelength
spectrum.
Figure 28: Spectral response of 1-stage interleaver before tuning
Figure 29: Spectral response of 1-stage interleaver after tuning
42
1543 1544 1545 1546 1547 1548 1549 1550 1551-50
-48
-46
-44
-42
-40
Pow
er (d
Bm)
Wavelength (nm)
3.2.6 Tuning of 2-stage MZI Based Wavelength Interleaver Fused fiber couplers of splitting ratios 45:55, 35:65 and 5:95 which were fabricated in the
lab were used. But later, the second coupler was replaced with 30:70 coupler that was
available in the lab (procured from Renka corp.). The first two couplers were initially
concatenated to realize a single-stage MZI. The initial FSR was measured, and the
differential delay was determined to be present in the upper arm.
Then we concatenated the third coupler (5:95) to the single-stage MZI to realize a two-
stage MZI. The initial output spectrum of the 2-stage MZI was recorded on the OSA. By
following the procedure, described previously for estimating the initial differential path
length in the second stage, we found that the upper arm of the second stage was longer
than the lower arm.
Figure 30: Spectral response of 2-stage interleaver during the process of tuning
As described earlier, since the splitting ratio of the second coupler is 30:70, the required
differential delay to achieve flattop response is 12 2 LL ΔΔ =′ , and it should be inserted in the
lower arm. Following the technique described in section (3.2.3) for tuning a single-stage
MZI, the lower arm of the second stage of the 2-stage MZI was elongated. The spectrum
obtained after partially tuning the second delay is shown in Fig. 31. The interleaver was
found to have 3.4 dB to 3.6 dB loss and peak-to-peak extinction ratio of 10dB.
43
1549.32 1550.12 1550.92 1551.72-54
-52
-50
-48
-46
-44
-42
Pow
er (d
Bm)
Wavelength (nm)
Figure 31: Spectral response of partially tuned 2-stage interleaver
3.2.7 Discussion The spectral response was found to be quite unstable when the fiber was exposed to the
flame. Hence online monitoring was not possible while adjusting differential paths of
both MZI stages. Hence, length was first increased, then, after the flame was withdrawn,
once the spectrum was stable, measurements were made. Therefore it was not possible to
tune exactly with the precision the technique allows. The peak-to-peak extinction ratio of
the interleaver was seen to increase with the FSR; this extinction ratio could be increased
further by a proper choice of splitting ratios of the couplers. The wavelength resolution
limit of the OSA also limits a precise measurement of the actual extinction ratio. The
minimum possible change in the FSR during tuning of the MZI by this technique was
difficult to estimate because of the following reasons: the smallest real-time change in the
FSR during tuning was much less than the resolution of the OSA. Further, the effect of
this small change in FSR was difficult to record in real-time during the tuning process
because of short-duration small-scale real-time fluctuations that occur in the online
response of the MZI.
44
Chapter 4
Y-Junction Flat-Top Interleaver 4.1 Introduction Interleavers play an important role in increasing the data carrying capacity of the
networks as the demand for the same has been on the rise. Interleavers exist in various
forms which use different principles or technology, using the optical fiber. But all of
these suffer from lack of compactness and ruggedness. Planar Lightwave Circuits (PLC)
offer a workable solution to these shortcomings. Polymer based waveguides have
attracted much attention of late due to their tunability over a wide range of wavelength
and wide choice of materials with varied properties. Recently, Qiang Wu, et al.,[2] have
reported polymer optical waveguide interleaver using Y junctions. They have replaced
directional couplers by Y junction components as they have following advantages:
1. Ease of fabrication
2. Wavelength insensitivity leading to much larger overall bandwidth of the
interleaver
3. Compact length due to smaller interaction length
4.2 Principle of Operation The structure consists of five Y junctions forming a cascade of two MZI’s similar to the
flat-top interleaver in the all-fiber form. The direction couplers are replaced by Y
junctions as shown in Fig. 23. The arms in between the Y-junctions are of unequal
lengths and form unbalanced MZI’s.
The first Y-junction (Y1) is a single-mode Y-junction, i.e., all of its branches support only
a fundamental mode. Here, its function is as a 3dB power splitter. The remaining Y-
45
junctions are of similar structure with their branches and stems being of different
dimensions. Their function is to convert, split and combine modes.
Figure 32: Schematic diagram of Y-junction based 2-stage MZI interleaver
Consider field E1 of amplitude A input at Y1. The two signals arriving at Y-junction differ
by a phase β' ΔL1 corresponding to path difference of ΔL1, where β' is the
propagation constant of each Mach-Zehnder arm.
At Y2 the fields entering the junction would be [2]
E1 = A cos α exp (iβ' ΔL1/2)
E2 = A sin α exp (-iβ' ΔL1/2)
These fields together create two lowest order propagating modes in Y2 because this
junction has a larger V number and is such that it supports these modes. The fields of the
two lowest-order local normal modes in S are the same E1/√2 with propagation constants
β0 and β1 respectively. Y3, Y4, Y5 are structurally identical to Y2. Let Δβ be the
difference in the propagation constants of these two modes. It is also noted that the
branches of the junctions support the fundamental mode only. Y3 splits the two modes in
the main arm into two single modes in the branches at its output end. Field at the upper
output branch of Y3 due to field E1 and E2 is given by
E3 = E1 cos(1/2 ΔβS1) − i E2 sin(1/2 ΔβS1)
Similarly, field at the lower output branch of Y3 due to field E1 and E2 is given by
E4 = − i E1 sin(1/2 ΔβS1) + E2 cos(1/2 ΔβS1)
This shows that the combination of Y-junctions in the above manner acts as a directional
coupler [2].
Using the same logic as in the case of all-fiber flat-top interleaver it can be shown that
Y5 Y4Y3Y2 Y1
ΔL1 ΔL2
S1S2
(4.4)
(4.3)
(4.2)
(4.1)
46
λλΔ
=Δeffn
L2
2
1
The two signals arriving at Y4 differ by a phase β' ΔL2 corresponding to path difference
of ΔL2. these fields can be expressed as
E5 = E3 exp (iβ' ΔL2/2)
E6 = E4 exp (-iβ' ΔL2/2)
Y4 and Y5 together act as another directional coupler. Thus the fields at the ouput
branches of the last junction are given by
E7 = E5 cos(1/2 ΔβS2) − i E6 sin(1/2 ΔβS2)
E8 = − i E5 sin(1/2 ΔβS2) + E6 cos(1/2 ΔβS2)
This resulting power at the upper output branch of the last junction is