`
`Photonic add-drop multiplexing perspective for next generation optical
`networks
`
`Benjamin B. Dingel* and Achyut Dutta**
`*Corning Incorporated, Optical Network Equipment Research Dept.
`**Fujitsu Compound Semiconductor, Inc. Lightwave Laboratory
`
`ABSTRACT
`An add/drop multiplexer (ADM) is recognized as one of the basic building blocks to extend DWDM networks from a static
`point-to-point system into a next generation, dynamic, re-configurable, programmable optical network. The objective of this
`paper is twofold: (1) provide an extensive overview tutorial of the numerous existing implementations of ADM, (2)
`categorize these different ADM implementations, and (3) assess their respective limitations and impacts on an evolving
`optical network. Toward this goal, a clear distinction between an OADM and a WADM node is made in terms of add/drop
`port configuration using a functional black box approach, and the types of component and device technologies that support
`these structures. Our second objective is to use this classification scheme to project, what we believe, is the functional form
`of the next generation ADM module. This is accomplished by taking into account major trends and developments in the
`optical networking arena. Lastly, some technical perspectives and directions toward the form of the next generation ADM are
`presented.
`
`Keywords: add and drop multiplexer, optical network,
`
`1.0 INTRODUCTION
`
`An add/drop multiplexer (ADM) is an essential and enabling device in the rapidly developing dense wavelength
`division multiplexing (DWDM) optical fiber communication networks [1]. ADM is recognized as one of the basic building
`blocks to extend the DWDM network from a static point-to-point system into a next generation, dynamic, re-configurable,
`programmable optical network. ADM provides, in its very fundamental form, the function of selectively removing and adding
`a number of wavelength channels without disrupting the other propagating wavelength channels being carried by a DWDM
`transport line. The dropped channels are then available either for regeneration purposes or access by client terminals.
`In technical and commercial literature, ADM is commonly referred to by many different terms. Two such familiar
`terms are Optical Add and Drop Multiplexer (OADM) and Wavelength Add and Drop Multiplexer (WADM). Often, these
`two terms are used interchangeably. Moreover, ADM is oftentimes discussed in two different functional contexts or levels
`without any clear clarification. These two contexts are either as (a) optical module, on one hand, and (b) network node, on the
`other hand. Although this vagueness does not lead to actual technical difficulties, it sometimes introduces confusion. In this
`paper, we will differentiate OADM from WADM, and state clearly the context in which they are used. We will also
`concentrate more on the dynamic implementations of ADM rather than static ADM.
`
`1.1 Functional context of ADM:
`Fig. 1 shows the interplay of three important levels related to ADM namely: (1) ADM as an optical module, (2)
`ADM as a network element or node, and (3) ADM node within an optical ring network. ADM-as-optical-module is the
`central element in any ADM network node. At this level, there are numerous reported implementations to obtain the basic
`add and drop function of a selected wavelength channel. Generally speaking, the particular choice of implementation greatly
`determines the overall architecture and performance of the ADM as a network node. As shown in the upper portion in Fig. 1,
`ADM-as-network-node is generally positioned between two network terminals or nodes or at any intermediate location
`within the optical ring networks where local access to only a fraction of the propagating wavelength channels is required.
`In order to maximize the performance of the optical network, it is desirable that ADM possesses both (1) excellent
`optical module properties, and (2) well engineered network node characteristics. At the module level, the optical properties of
`ADM should approximate the following ideal features namely: (1.1) unity transmission, flat-top, nearly zero-ripple frequency
`response over the pass-through spectrum, (1.2) infinite rejection ratio over the drop spectrum, (1.3) sharp roll-off or steepness
`in between the add and drop spectrum, (1.4) negligible phase dispersion on the pass-through or drop wavelengths, (1.5) high
`isolation of all pass-through wavelength channels from transient negative effects of a tuning filter (also known as (cid:147)hitless(cid:148)) ,
`(1.6) low insertion loss, and (1.7) negligible polarization mode dispersion (PMD) & polarization dependent loss (PDL)
`values. Deviations from these ideal properties impact the quality of the output optical signals leaving an ADM network node.
`
`394
`
`Active and Passive Optical Components for WDM Communication, Achyut K. Dutta,
`Abdul Ahad S. Awwal, Niloy K. Dutta, Katsunari Okamoto, Editors, Proceedings of SPIE
`Vol. 4532 (2001) © 2001 SPIE · 0277-786X/01/$15.00
`
`Capella 2015
`Fujitsu v. Capella
`IPR2015-00726
`
`
`
`As a network node, ADM should be designed not only to provide basic add-and-drop capability but also to deliver
`other essential node characteristics such as; (2.1) scalability in number of drop wavelength channels and bit rate, (2.2)
`modularity, (2.3) efficient wavelength utilization, (2.4) optical performance monitoring, (2.5) fault detection and isolation,
`(2.6) optical amplification, (2.7) channel power equalization, (2.8) optical layer protection, (2.9) redundancy, and (2.10)
`service restoration. Besides these optical node characteristics, ADM node might also contain electrical node functionality
`such as (2.11) 2R/3R regeneration, (2.12) bit rate adaptation, and (2.13) grooming. All these requirements should be met at
`
`[3] ADM Node within Interconnected Optical Ring
`Networks
`
`OXC
`
`Client Interface
`Module
`
`Rack
`
`[2] ADM as Network
`Node
`
`Amplifier
`
`[1] ADM as
`Module
`
`In
`
`Add
`
`Optical Circuit
`
`Electronic Parts
`
`Module Controller
`
`Out
`
`Drop
`
`Fig. 1. ADM-as-optical-module is part of ADM network node which is positioned between two
`other network nodes of interconnected optical ring networks.
`the minimum cost while maintaining high quality signal. These node characteristics or requirements have dual consequences.
`First, they constraint the design of ADM-as-module. Secondly, they define in large part the optical ring network(cid:146)s
`survivability, quality of service, and over-all performance.
`At the network level, the topology of the actual optical networks and network upgrade path designs, (among other
`things such as economic viability, etc) put a general framework on which these ADM nodes and ADM modules must be
`designed and optimized. Ideally, this framework can be considered from day one, but in practice, the overall network is still
`evolving and its (cid:147)evolutionary form(cid:148) is still not yet clear. The need to continuously track this network trend, meet new
`demands and respond appropriately lead the telecom industry to introduce, build and deploy next generation ADM nodes or
`next generation ADM modules.
`
`1.2 Functional relationship between ADM and OXC
` Although this paper is focused on ADM, it is important to point out the functional relationship between ADM and
`cross-connect (XC). As shown in the upper portion of Fig. 1, an example of optical network of the kind that we are
`
`Proc. SPIE Vol. 4532
`
`395
`
`
`
`considering consists of two optical fiber rings which are interconnected by a network node commonly known as XC.
`Functionally speaking, a XC routes or transfers any optical signals coming from any input fiber ports of either optical ring to
`any output fiber ports of either ring. This interconnection is not limited to optical rings but can be extended easily to mesh
`topology with the XC having greater degree of connectivity. Furthermore, XC can, in general, also provide add/drop
`functionality for all connected fibers. XC is considered functionally more general than ADM. XC and ADM are generally
`regarded as two distinct network nodes. But in principle, ADM is the simplest form of XC having one degree of node
`connectivity.
`As the optical network topology evolves from simple ring to multiple interconnected rings, and finally to mesh, a
`corresponding evolution on ADM will also occur. In this paper, we will also touch on the functional evolution of ADM from
`a typical ADM node towards a more sophisticated ADM-based XC.
`
`1.3 Objectives of the Paper
`In order to have a clear understanding of present ADM nodes, possible architectures of the next generation ADM
`nodes, and future functional relationship between ADM and OXC, it is essential to have a comprehensive overview of the
`different existing implementations of ADM-as-module. However, due to the enormous number of reported ADM module
`designs and a lack of clear classification method, it is oftentimes difficult to assess their respective limitations and impacts on
`the ADM node, and eventually to the evolving optical transparent network. In this paper, our objective is twofold. First, we
`present an extensive overview tutorial of the different existing implementations of ADM modules, and categorize them using
`a new, simple but powerful classification scheme. Toward this goal, a clear distinction between an OADM from a WADM is
`made in terms of (1) add/drop port configuration using a functional black box approach and (2) the types of component and
`device technologies that support these structures. Our second objective is to use this classification scheme to project, what we
`believe, is the functional form of the next generation ADM module. This is accomplished by taking into account major trends
`and developments in the optical networking arena. We will refer to this future ADM module as Photonic ADM (PADM). The
`relationship of the OXC and PADM will then be evident from the discussion in the paper.
`This paper is organized as follows. First, we introduce our classification scheme in section 2.0. Then we present
`the hierarchy of an ADM. The comparison and discussion of existing OADM implementations are given in section 3.0. The
`corresponding discussion for WADM implementations is given in section 4.0. We offer our perspective on the next
`generation ADMs with an explanation of PADM in section 5.0.
`
`2.0. GENERAL CLASSIFICATION OF ADM
`
`The numerous existing and proposed optical module-based ADMs can be classified in many different ways.
`Generally speaking, they have been classified according to whether they are (a) static or dynamic ADMs, (b) serial-based or
`parallel-based ADMs, (c) single-function- or multiple-function-based ADMs, (d) add/drop wavelength count capability (high,
`moderate, or low), and (e) technology specific-ADMs.
`In this paper, these different ADM implementations are classified rather uniquely and unconventionally. ADMs are
`classified based on their add/drop port physical configuration. The port configuration of any ADM will be defined by the
`following convention, (N x L) x 2D where (N x L) represents N input ports, L output ports that are designed to accept only
`multiplexed signals, and D denotes add-and-drop port pairs that are engineered to accept only per-λ signals or wavelength-
`banded signals or both. When D is equal to zero, the ADM is simply referred to as a (N x L) module. In order to add M
`numbers of wavelengths that an ADM module is designed to operate, the notation will be refined to a M-wavelength channels
`x (N x L) optical module.
`Using this classification, the numerous ADM implementations can be grouped into three general classes as shown
`in Fig. 2. Within the context and confine of this paper, we will refer to the first class and second class of ADMs as OADM,
`and WADM, respectively. The third class will be called PADM. From our discussion, it will become more apparent why
`these labels are appropriate. As shown in Fig. 2, the OADM has M-wavelength=channels x (2 x 2) port configuration where
`(2x2) stands for one input port, one output port, one multiplexed add port, and one multiplexed drop port. Functionally, it
`accepts multi-wavelength input signals at the main input port, selects S number of discrete, arbitrary wavelength channels,
`directs them to the primary drop port, adds new S number of discrete wavelength channels through a primary add port, and
`finally recombines the resultant multi-wavelength signals before they exit the ADM at the main output port.
`On the other hand, a WADM is configured as a M-wavelength=channels x (1 x 1) x 2K optical module where (1x1)
`stands for one main input port, one main output port and 2K number of secondary add and drop ports. These 2K ports are
`meant to individually access the K add/drop per-wavelength channels. The add/drop ports have pre-determined, fixed
`wavelength assignments. Note that an OADM can be transformed functionally to a WADM by two different approaches. The
`simplest way is to connect external demultiplexers (Demuxes) and multiplexers (Muxes) to the drop and add ports of OADM,
`
`396
`
`Proc. SPIE Vol. 4532
`
`
`
`respectively. Likewise WADM can be converted to OADM in the same manner. The advantages and disadvantages of doing
`this approach will be discussed in section 4.0.
`While OADM and WADM represent most existing and proposed ADM optical modules, the third class is what we
`believe to be the (cid:147)functional form(cid:148) of the next generation ADM. PADM is a combination of OADM and WADM with some
`extra new capabilities. It is a M-wavelength=channels x (2 x 2) x 2(K+B) optical module where (2x2) stands for the same
`structure as OADM, and the 2K stands for the numbers of secondary add and drop ports just like the WADM. However
`unlike the WADM, these secondary ports take in wavelength-banded signals (which are represented by B) as well as per-
`wavelength signals (which are represented by K). The thick slash lines in the input / output arrows represent wavelength-
`multiplexed signals while the thin slash lines in the secondary ports denote wavelength-banded signal. Wavelength-banding
`consists of 4 or more adjacent wavelength channels per band. Functionally, PADM-as-module has expanded capability
`compared with conventional OADM and WADM. Table 1 provides the general features of these three classes of ADMs. A
`more detailed discussion of PADM is given in section 5. As a network node, ADM would contain two or more ADM-as-
`optical-modules to provide dual directions for the east and west transmissions. Note that Fig. 2 shows only unidirectional
`transmission and extension to bi-direction propagation would be straightforward.
`The distinction between OADM, WADM and PADM has many significant beneficial consequences. For one, it
`provides a general platform to classify, segment, and compare the numerous existing technical implementations of ADM in a
`coherent and comprehensive way. Secondly, it identifies the respective applications of the different ADM implementations
`based on their technical merits and demerits. Lastly, it would help us to understand the current technical trend and gain
`insight to the next generation ADMs.
`
`M-λλλλ
`
`s-λλλλ(cid:146)
`M-λλλλ
`
`Input
`Port
`Add
`Port
`
`Input
`Port
`
`OADM
`
`WADM
`
`
`
`(M-s)λλλλ==++++====s-λλλλ(cid:146)=
`Output Port
`
`Drop Port
`
`s-λλλλ
`
`
`(M-s)λλλλ==++++====s-λλλλ(cid:146)=
`Output Port
`
`(a)
`
`(b)
`
`K-add Ports K-drop Ports
`
`
`(M-s)λλλλ==++++====s-λλλλ(cid:146)=
`M-λλλλ
`Output Port
`s-λλλλ
`Drop Port
`B-drop
`B-add
`K-drop
`K-add
`Ports
`Ports
`Ports
` Ports
`Fig. 2. Three different groupings of add and drop multiplexer (ADM as optical module) according to their functional port
`configurations. Photonic ADM (PADM) is a combination of both optical ADM (OADM) and wavelength ADM (WADM)
`with new and extended functionality. S-λ(cid:146)=and S-λ represent the number of add and drop wavelengths, respectively. The thick
`slash lines in the input / output arrows (in all ADMs) represent wavelength-multiplexed signals while the thin slash lines in
`(PADM only) secondary ports represent wavelength-banded signals. Only one direction of the transmission is depicted.
`
`(c)
`
`Input
`Port
`s-λλλλ(cid:146)
`Add
`Port
`
`PADM
`
`Add and Drop Multiplexer (ADM)
`
`Classifications of
`
`3.0 HIERARCHY OF ADM
`
`With this classification, OADM, WADM and PADM can be subdivided further into smaller groupings to establish
`a form of hierarchy of ADMs. Here we will concentrate more on the dynamic implementations of ADMs rather than static
`ADMs. As shown in Fig. 3, this ADM hierarchy consists of 4 layers. The first layer is composed of OADM, WADM and
`
`Proc. SPIE Vol. 4532
`
`397
`
`==
`====
`==
`=
`==
`====
`==
`=
`==
`====
`==
`=
`
`
`PADM groupings. Again, the distinction between these classes is based on the physical port configuration. In the second
`layer, the different OADM and WADM implementations are categorized broadly into two major groupings based on the
`physical mechanism involved in the dynamic selection of wavelength(s) to be dropped or added. These two major groupings
`are called (1) (cid:147)tuning-based λ-selection(cid:148), and (2) (cid:147)switching-based λ-selection(cid:148). ADM implementations under these two
`broad groupings can be segmented further into smaller sub-groupings or subsets according to the generic optical structures
`employed. This is the third layer. The last or fourth layer is based on the actual design implementations.
` The detailed discussion of this hierarchy starts with an OADM in section 3.1 and is followed by a WADM in
`section 4.0. For an OADM, this hierarchy is broken down into a (cid:147)tuning-based OADM(cid:148) and (cid:147)switching-based OADM(cid:148). We
`will evaluate these implementations purely on the following issues: (1) level of functionality, (2) component counts for multi-
`channel operation and configurability, and (3) scalability. Although the technical performance will not be evaluated in detail,
`we will pin point key parameters that limit their performance. Brief explanation of these different (cid:147)tuning based OADM(cid:148)
`designs are given below. Furthermore, these are briefly summarized and tabulated in Table 2.
`Features
`OADM
`WADM
`PADM
`Port configuration
`(2x2)
`(1x1) x 2S
`(2x2) x 2S
`No. of λ-channel
`M
`M
`M
`Add / Drop Ports
`1-add, 1-drop ports
`1-add, 1-drop ports
`none
`none
`S-add, S-drop ports S-add, S-drop ports
`
`Comments
`S stands for secondary
`wavelength add/drop ports
`
`Primary
`Secondary
` Access Granuality
`
`multiplexed =λ
`per-λ
`
` λ− banding
`
`yes
`none
`
`none
`
`none
`yes
`
`none
`
`Primary ports only Secondary ports only
`
`yes
`yes
`
`yes
`Both primary and
`secondary ports
`
`3 levels of granuality is
`required for PADM
`
`flexibility demands dual
`access for PADM
`
`Fixed
`
`Variable
`
`none
`
`none
`
`yes
`
`none
`Application
`depended
`
`yes
`
`supported
`Application
`depended
`
`client configurability is
`required for PADM
`
`Transformation of OADM to WADM (and vice versa)
`2 external Demux /
`2 external Demux /
`Mux devices
`Mux devices
`
`functionally build-in in
`PADM
`
` multiplexing
`penalty
`
`yes
`
`none
`Concatenation of OADM units
`
`Access path to add/drop
`K-port vs λ=assignment
`
`No. of Secondary Ports
`
`Approach # 1
`(requirements)
`
`Approach # 2
`(requirements)
`
`"Drop and terminate" functionality
`
`"Drop and continue" functionality
`Optical Layer Protection
`
`Dedicated protection
`
`Shared protection
`
`Network Application
`
`Additional Functionality
`
`Well-suited
`
`Suited
`
`yes
`
`yes
`
`Good for direct client
`terminal connection
`Good for ring
`interconnection
`
`Need to be engineered at node level
`Need to be engineered at node level
`Simple Ring,
`Multiple
`Interconnected Ring,
`and Mesh
`
`Interconnected
`Optical Ring
`
`Simple Optical Ring
`with direct Client
`interface
`
`build-in
`
`PADM is very appropriate for
`network evolution because of its
`flexibility
`
`Table 1. General features of the three different groupings of ADM. Note that wavelength-banding, client configurability,
`and dual access (primary and secondary ports) to add/drop signal are some of the new requirements for PADM.
`
`3.1 Tuning-based OADM
`Tuning-based OADM depends on two fundamental elements: (1) tunable filters to select a particular wavelength(s)
`
`398
`
`Proc. SPIE Vol. 4532
`
`
`
`from a bundle of discrete, multiplexed wavelengths, and (2) optical structures to discriminate, isolate, and direct the selected
`wavelength to a drop port. Commonly used tunable filters in OADM are Fiber Bragg Grating (FBG) [2-3], Fabry-Perot
`interferometer (FPI) [4-6], ring resonator (RR) [7-10], Michelson-Gires-Tournois interferometer (MGTI) [11-14], dielectric
`thin film filter (TFF) [15], acousto-optic filter (AOTF) [16-18], and MEMS-based FPI [19-22]. These tunable filters can be
`constructed in all-fiber, liquid crystal, planar, MEMS or bulk optics forms.
`In all tuning-based OADM implementations, configurability is accomplished by tuning a filter to a different
`assigned wavelength. Some common methods of tuning these filters are based on inducing optical path length change in the
`spacing of the grating or filter or resonator either through thermal, piezoelectric, electro-optic, and mechanical induction.
`Depending on the filter technology employed, there are trade-offs related to tuning speeds, available tuning ranges, crosstalk,
`insertion loss, cascade-ability, size or dimension, and cost. Thermal tuning is associated with relatively slow switching speeds
`and limited tuning ranges. Mechanical tuning, on the other hand, provides wide tuning ranges and moderate speed.
` With regards to the optical structure, there are at least 4 generic optical configurations that can be used to
`differentiate various implementations. These structures are (1) Mach-Zenhder Interferometer (MZI)-based, (2) directional
`coupler (DC)-based, (3) circulator-based, and (4) acousto-optics (AO)-based configurations. The central element in any of
`these implementations is a tunable filter.
`The discussion of different optical structures will have the following format. First, a general description of the
`structures and basic operations will be given. Then, the multi-channel operations of the particular ADM will be discussed.
`Lastly, the advantages and disadvantages of the structures will be explained. Due to the limited space, these issues will be
`
`Add / Drop Multiplexer (ADM)
`
`What is port
`configuration?
`
`What is
`physical
`mechanism for
`λλλλ-selection?
`
`OADM
`
`WADM
`
`PADM
`
`Tuning based
`λλλλ-selection
`
`Switching based
`λλλλ-selection
`
`Switching based
`λλλλ-selection
`
`Tuning based
`λλλλ-selection
`
`MZI-
`based
`
`Circulator-
`based
`
`DC-
`based
`
`AO-
`based
`
`Serial-
`based
`
`Parallel-
`based
`
`AWG -
`based
`
`Non - AWG
`based
`
`Concatenated
`ADM
`
`What is
`the key
`optical
`structure?
`What are the actual design implementations ? ----> see Table 2
`Tunable FBG, AO, TF etalon ,
`Low loss 2x2 switch array ,
`Critical
`ring resonator,
`technologies
`High=λ-count, low loss AWG
`Low cost, low loss circulator
`suitable for high number of
`channels to be added/drop
`
`Add/drop
`percentage
`
`General
`Comment
`
`suitable for low number of
`channels to be added/dropped
`Circulator price needs to go down,
`Circulator crosstalk needs
`improvement,
`
`Low loss 2x2
`discrete switches ,
`FBG, DC
`
`moderate number of
`channels to be added
`
`array of 2x2 switches is essential,
`integrated SW array - AWG is promising
`
`Fig. 3. The 4-layer hierarchy which classifies numerous implementations of ADM based on port configuration,=λ-
`selection mechanism, and generic optical structures.
`explained only briefly. Readers are encouraged to refer to the references for further information.
`
`3.1.1 MZI-based tunable OADM
`The first basic optical structure is referred to as Mach-Zenhder Interferometer (MZI)-based OADM since it uses
`two-beam interference as the basic physical mechanism for both wavelength filtering and spatial discrimination. The main
`optical configuration consists of a MZI with an optical filter in one or both arms of the interferometer. Variation of this
`implementation depends on the types of optical filter employed and the path arm length difference condition of the MZI.
`
`Proc. SPIE Vol. 4532
`
`399
`
`
`
`Classes of ADM
`
`Actual
`Implementation
`Tuning-based₧₧λλλλ- selection
`
` Optical Structure
`
`OADM
`
` (1) MZI-based
`
`MZI+2 FBG
`
`MZI+2 FPI
`
`MZI+ Ring Resonator
`(RR)
`Assymmetric MZI
`Cascaded MZI
`
`Key
`Technology
`
`Multi-λλλλ =
`operation
`
`Special
`Features
`
` Major Issues/
`limitations
`
`Ref.
`
`FBG, Ring
`Resonator
`
`cascading
`multiple units
`
`compact
`
`requires ∆L = 0, perfect
`50:50 power spliters
`
`23-32
`
`multi-function
`
`fabrication tech. for RR
`is not yet mature
`
`33
`
`MZI
`
`cascade
`
`well understood
`approach
`
`phase instability of MZI
`
` (2) DC-based
`
`DC+ FBG
`
`DC + FBG
`
`cascade of
`multiple units
`
`very simple
`structure
`
` (3) Circulator-based Circ. + FBG + Circ.
`
`Circ. + FPI + Circ.
`
`Circ. + MGTI + Circ.
`
`Circulator,
`FBG.Ring
`Resonator, MGTI
`
` inserting series
`of filters
`between the two
`circulators
`
` (4) AO-based
`
`AOFT-OADM
`
`AOTF
`
`built-in
`
` (5) Ring Resonator-
`based
`
`Switching-based₧λλλλ- selection
` Optical Structure
`
`Ring Resonator
`
`simple
`implementation
`
`high IL, requires
`circulator with low
`crosstalk
`
`multi-function
`
`compact, new
`approach
`
`Stability and tolerance
`of MGTI
`high IL, requires high
`power RF signals
`fabrication tech. not
`mature
`
`65-68,
`70-71
`
`44-50
`
`36-40
`
`35
`
`41
`
`51-54
`
`7,8-10
`
` (1) Parallel-based
`
` (2) Serial-based
`
`2x (Demux+ Parallel
`Sw Array + Demux)
`Circ. + (1xN) Sw +
`FBG Array+ (N+1) Sw
`+ Circ.
`Circ. + series of (1x2
`Sw + FBG) + Circ.
`
`AWG
`
`built-in
`
`full
`configurability
`
`bandwidth narrowing,
`uses 4 Demuxes
`
`55,73-
`75
`
`1xN Switch
`
`hitless,
`
`requires low loss (1xN)
`switch
`
`1x2 SW, FBG
`
`built-in
`
`hitless, signal
`need not pass all
`FBGs
`
`IL limited
`
`69
`
`72
`
`WADM
`
`Stackable, alternating
`(TF + Sw) array
`Tuning-based λλλλ- selection
` Optical Structure
` Concatenated
`OADM
`
`Series of (DC+FBG)
`
`Series of (MZI-based
`OADM)
`Series of (Circ.+FBG+
`Circ.)
`Spliter + Amplifiers +
`Broadcast, Select &
`Filters
`Combiner
`Switching-based λλλλ- selection
` Optical Structure
`
` (1) AWG-based
`
` (2) Demux-
`Combinerr-based
`
`AWG + Parallel Sw
`Array + AWG
`
`AWG in Folded
`Arrangement
`AWG in Loop-back
`Arrangement
`DC + AWG + Array of
`(Phase Shifter +
`Reflector)
`Demux + Parallel Sw
`Array + (Nx1) Coupler
`
`requires low loss, low
`phase dispersionTF
`
`DC + FBG
`
`FBG, Ring
`Resonator
`
`Circulator
`Optical
`Amplifieres (OA)
`, Filters
`
`cascade
`
`very simple
`structure
`
`cascade
`
`flexible
`expandability
`
`IL limited
`
`IL limited
`
`huge loss due to
`multiple circulators
`Noise (ASE) limited,
`requires lot od OA
`
`AWG, Array of
`2x2 SW
`
`built-in
`
`low component
`count, use 2
`AWGs
`use only 1
`(NXN) AWG
`use only 1 (NxN)
`AWG
`
`use only 1 (1xN)
`AWG only
`
`bandwidth narrowing
`
`55
`
`crosstalk limited
`
`crosstalk limited
`
`56-58
`
`requires power
`balancing 50:50 split in
`DC,
`
`59-63
`
`IL limited, bulky
`
`Table 2. Partial list of the different ADM implementations with their respective features and limitations. They
`are categorized using our classification scheme.
`Three such variations are (1) MZI with two similar fiber bragg gratings in both arms [23-32], (2) MZI with two similar
`
`400
`
`Proc. SPIE Vol. 4532
`
`==
`=
`₧₧
`₧
`₧₧
`
`
`Fabry-Perot interferometers (or dielectric thin film) in both arms, and (3) MZI with a ring resonator in one arm [33]. Slight
`variations of the same concepts include the use of polarization(cid:150)based MZI with tunable filters [34], use of other forms of
`gratings such as holographic film, and a Michelson interferometer set-up with filter(s)[33]. These variations will not be
`discussed here. In general, all these implementations together with above-mentioned variations can be summed up under the
`larger label of grating- or filter-assisted MZI-based OADM.
`Generally speaking, the basic operations involve injecting the multi-wavelength signal at an input port, splitting it
`at coupler #1 into two beams, changing the directions or phases of the two beams by propagating them through the tunable
`filter(s), and then recombining them at coupler #2 (at coupler #1 for the reflected beams) to interfere at a particular phase
`condition so that wavelength λ0 exits at drop port while the rest of the wavelengths leave the OADM at the main output
`port. The same operation holds true for the add wavelength λ0(cid:146) signal coming from the opposite direction or conjugate port.
` In order to obtain the multi-channel operation of this structure, the basic unit is cascaded with a number of other
`MZI-based OADM units having different wavelength assignments so that the overall combination leads to simultaneous
`multi-wavelength, independently configurable, add-and-drop capability. This implementation is best described as a WADM
`with a representative schematic shown in Fig. 4(a). The general advantage of this type of implementation is its simple
`structure, small size using integrated planar technology, and manufacturability. However, the technical performance is
`strongly negated if the conditions for (i) zero-path-length-difference (∆L=0), and (ii) perfect 50:50 power splitter/combiner of
`the MZI, are not satisfied. In practice, both these conditions are oftentimes difficult to perfectly satisfy especially with multi-
`wavelength input signals. These conditions lead to increased crosstalk.
`
`3.1.2 Circulator-based tunable OADM
`In the circulator-based grouping, the basic optical structure consists of a tunable reflective or blocking filter
`sandwiched between two optical circulators. A multi-wavelength light beam enters the left-hand optical circulator (refer here
`as circulator A) at port 1, passes port 2, and travels toward a tunable filter. The tunable filter reflects back a selected
`wavelength λ0 toward port 2 and then port 3 of the circulator A, which functions as a fiber drop port. On the other hand, the
`wavelength channels transmitted by the tunable filter (the pass-through wavelengths) enter the right-hand circulator (refer to
`as circulator B), and then exit at port 3 as the output signal. For the case of adding a signal, the same operation holds but at
`the opposite direction. Here, the new wavelength λ0(cid:146) signal enters at port 1 of the circulator B, passes through port 2,
`propagates toward the direction of the filter so that it is reflected back toward the port 2 and then leave at port 3 of circulator
`B together with the pass-through wavelength channels
` There are at least 3 different variations of this type of implementations depending on tunable optical filters. In
`this paper, these implementations are briefly designated as (1) circulator + FPI+ circulator [35], (2) circulator + FBG+
`circulator [36-40], and (3) circulator + MGTI+ circulator [41]. These notations are used to facilitate cross referencing with
`the information found in Table 2. The third variation is based on a recently proposed multi-function optical filter [42-43] by
`one of the authors. Unlike FBG and FPI, MGTI filter is a multifunction device that offers added features; better performance
`compared with conventional FPI and FBG filters. Furthermore, it also offers unique potential function of being able to
`provide high capability, multi-wavelength point-to-point transmission within fiber ring network.
`Converting this generic structure to perform multi-channel operation is very simple. We need only to insert
`tunable filters between these two circulators to provide simultaneous multi-wavelength, re-configurable, add-and-drop
`functionality as shown in Fig. 4(e). The main advantage of this design is that there is no zero-path-length-difference (∆L=0)
`condition to satisfy as in the case of MZI-based implementation. This translates into a high tolerance and rugged optical
`structure. On the other hand, the disadvantage is related to the higher insertion loss and crosstalk, expensive circulators, and
`limited level of scalability due to insertion losses.
`
`3.1.3 Directional coupler (DC)-based tunable OADM
`For the directional coupler-based grouping, the basic operation relies on different wavelength coupling
`mechanisms employed in directional couplers with filter or grating [44-50] assisting in the wavelength selection. These
`mechanisms include adiabatic coupling via mode conversion, coupling via supermode, grating-frustrated coupling, bragg-
`assisted multimode-interference(MMI) coupling, vertically stacked waveguide coupling, and others. A representative
`implementation is schematically depicted in Fig. 4(b). Planar and fiber-based designs have been demonstrated both in
`polished and tapered fibers. We will refer to this family of implementations simply as (cid:147)directional coupler + FBG(cid:148) which can
`be cross referenced with Table 2. Here, we will discuss only one fiber implementation [44]. It consists of one directional
`coupler with a stretched middle fiber