Fundamentals of Switches (Photonic Switching)

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 24

All Optical Switches/Couplers for communication

networks
1.1 Introduction

In optical communication systems information is transmitted in the form of modulated optical


signals from source nodes to destination nodes, which are interconnected by optical fibers.
Fiber optic lines offer advantages over other communication media in terms of better
bandwidth and reliability. Optical fibers offer much higher bandwidth than copper cables and
are more robust and less susceptible to various kinds of electromagnetic interference and other
undesirable effects [35]. The advantages of fiber optic cables over wire include greater
bandwidth over greater distances with reduced data loss and overall cost for long distance links
over time. Fiber optic cables are considerably less vulnerable than metal conductors to
unapproved “monitoring", RF problems, therefore suitable for telecommunications systems,
digital data links, LANs, and also finds application in illumination, medical and displays uses.

An optical telecommunication system demands a variety of devices for initializing,


transmitting, switching, amplifying, filtering, and processing optical signals. In a typical
optical communication system, switching systems also including repeaters and regenerators
reconnect the spans of optical fibers. These switching systems are used for routing signals to
destination and addition/removal of same from the optical fibers. In the optical
communication, an optical fiber forms a waveguide for the optical signal and thus generated
signal needed to switch from one waveguide to other. In analogous way to copper wire based
telecommunication system, fiber optic cable lines are interconnected to each other through
switches located at various places throughout the network.

1.2 All optical switch

According to basic definition, an optical switch enables signals in optical fibers or integrated
optical circuits to manage switching from one circuit to another precisely. Photonic switches
exploit the nonlinear material properties to perform and controls switching of light, irrespective
of how the light itself is switched. Theoretically, all optical switches are independent of bit rate
and protocols with unlimited scalability, which leads to more flexibility in the network [1]. An
optical switch directs an optical signal from one or more input ports to one or more output
ports. A basic 2×2 optical switch has two possible switching states, popularly termed as bar
and cross state (Figure 1.1).

(a) (b)
Figure 1.1 Basic 2×2 switch block with switching states (a) the "bar" state; (b) the "cross" state

Optical switches call for certain requirements though. As optical switches are usually in
operation at all times, this leads to unnecessary power consumption which is a major cause of
significant waste of resources so power consumption in optical switches must be minimized.
Optical switches are required to be space optimized so many channels can be accommodated in
a relatively small area, as scalability is a significant concern for optical switches. Almost all
All optical switches

fabrication methods are bound to have a significant dependence on the size, shape, and
structure of the devices, which explains the deep correlation between fiber optic switch design
and the physical design of optical fibers.

1.3 Applications for communication networks

In communication networks, fibre optic switch has been in use for many applications such as
routing, multiplexing, signal processing, distributed sensing, and optical logical and control
operations. There exists a wide variety of optical switches, using an equally wide range of
switching technologies, for numerous applications, largely related to telecommunication
networks, like network protection, restoration, monitoring and device redundancy. The
different applications require different number of ports, wavelength and polarization
dependence and switching times. Commercial applications for optical switches include
medical, aerospace, national defense, and other manufacturing industries. The following
applications are mainly considered in Optical Switching by Georgios I. Papadimitriou et al.
published in 2007 [1].

 Optical Cross-Connects: One of the most important applications of optical switches is to


provide light paths from any one of the several input nodes to the desired output node. A light
path [36] is a dedicated connection between two network nodes for an optical signal of fixed
wavelength. For this purpose, the switches are used inside optical cross-connects (OXCs) to
dynamically reconfigure them to support new light paths. Figure 1.2 shows an opaque network,
in which DWDM systems and line terminals are connected through an OXC, which is the base
for routing optical signals in the network, while OXCs dynamically make provisions for and
optimize transmission data paths. When the signals are needed to be switched on a packet-by-
packet basis, OXCs can be used to build packet switching networks. For packet switching, a
switching matrix with port count size of more than 100 with a switching time of around 10ns is
necessary.

Figure 1.2 The optical cross-connect in the opaque network; ref. [37].

 Protection Switching: Protection switching is a safety measure, allowing the completion of


traffic transmission even in the event of system or network-level errors i.e. to switch the traffic
stream from a primary fiber onto another idle redundant fiber, in case the primary fiber fails.
The optical protection switching usually requires optical switches with smaller port counts of 1
× 2 or 2 × 2 and the required switching time is in 1–10ms range. Switches employed for
protection switching have to be extremely reliable, since often these switches are single points
of failure in the network, and their breakdown will result in transmission data loss. Figure 1.3
represents the use of 1×2 switches for line protection applications. It is stated at [37] that, the
optical line protection switching in the optical domain will exist next to the protection
capability provided by the SONET/SDH layer, as it can also provide protection capabilities for
transmission protocols.
All optical switches

Figure 1.3 Line protection using two 1×2 optical switches, ref. [37]

 Optical Add/Drop Multiplexing: Optical add/drop multiplexers (OADMs) present in


network nodes are capable of inserting (add) or extracting (drop) optical channels
(wavelengths) to or from the optical transmission stream [38]. Using an OADM, channels in a
multiwavelength signal can be added to the traffic stream or dropped without any electronic
processing. Figure 1.4 depicts application of wavelength selective switches to add and drop
optical signals in a link. Wavelength selective switches can perform this operation in
accordance to wavelength of optical signals.

Figure 1.4 Add-drop applications using wavelength selective switches

 Optical Signal Monitoring: Monitoring of optical signal levels and parameters sometimes
referred to as optical spectral management of the network. Optical signal monitoring (OSM) is
a three step process. In the first step, it receives a small portion of optically tapped aggregated
WDM signal. In the next step, this signal is segregated into its individual wavelengths. And
finally, each channel’s optical spectra are checked for wavelength accuracy, optical CT and
power levels. The optical switch require to perform this operation is usually selected on the
basis of the system wavelength density, wavelength range of interest and the desired
monitoring strictness.

 Network Provisioning: In communication networks, this operation is required, when new


data routes have to be defined or existing routes need to be reconfigured. A network switch
should carry out reconfiguration requests over time intervals typically of the order of 1–10ms.
In provisioning of lightpaths, switches with port count size of more than 1000 are used inside
wavelength cross-connectors to reconfigure them to support new lightpaths.

1.4 Switch performance parameters

There are many important parameters which can be used to characterize suitability of a switch
for optical networks. The most important parameter of a switch is the switching time. Different
applications have different switching time requirements. Other important performance
parameters are summarized below, which are mainly suggested in Optical Switching by
Georgios I. Papadimitriou et al. published in 2007 [1].

 Insertion loss (I.L.): It is related to the port to port power transmittance. This is defined
as the fraction of signal power lost due to transmission through the switch and it
consists of coupling loss, waveguide propagation loss and other losses. This loss is
usually measured in decibels and must be as small as possible, ideally 0 dB for a
lossless path. In addition, the insertion loss of a switch should be almost same for all
input–output connections (loss uniformity).
All optical switches

 Excess loss (E.L.): It refers to the ratio of sum of all output power to the launched input
power. Since it quantifies from the ideal case where this ratio should be unity, therefore
E.L. should be as small as possible
  jPj 
Pex  dB =  10 log  
 Pi 
Where Pj is the output power at port Pj and Pi is the input power

 Crosstalk (CT): It refers to a broad class of phenomena generally defined as one in


which a signal transmitted on one circuit or channel or one part of a transmission
system creates an undesired impact on another circuit, channel or part, usually a
neighboring or connected channel. The chief causes of crosstalk include undesired
capacitive, inductive, or conductive coupling from one circuit, part of a circuit, or
channel to another. This is the ratio of the power at a specific output from the desired
input to the power from all other inputs. Low CT and high extinction ratio is usually an
indication of small signal interference or high signal quality

 Extinction ratio (ON–OFF switches): This is the ratio of the output power in the on
state of the switch to the output power in the off-state. This ratio should be as large as
possible. In telecommunications, extinction ratio (re) is the ratio of two optical power
levels, of a digital signal generated by an optical source, e.g., a laser diode, where P1 is
the optical power level generated when the light source is "on" and P2 is the power
P
level generated when the light source is "off" [ re  1 ]. The extinction ratio may be
P2
expressed as a fraction or in dB.

 Polarization and wavelength dependent loss (PDL & WDL): If the loss of the switch is
not equal for both states of polarization (TE and TM polarization modes) of the optical
signal, the switch is said to have polarization dependent losses [1]. In simple definition,
the polarization dependent loss is the ratio of the maximum and the minimum
transmission of an optical device with respect to all polarization states. Higher these
losses, more complex it is to monitor and compensate dynamically. Losses should also
be not much dependent on wavelength on input signal at least within the band of
interest.

Throughout the thesis work, we shall be using following definitions [39] to calculate the switch
performance parameters for selected optical path from 1 to 2 (bar state) and 1 to 3 (cross state)
as shown in figure 1.5 (a-b).

(a) (b)
Figure 1.5 Loss definitions for different optical path (a) bar state and (b) cross state
All optical switches

 P2  P3 
(a) Excess loss: Pex  dB   10log
P1

P  P 
(b) Insertion loss: I.L12  dB   10log  2  and I.L13  dB  10 log  3 
 P1   P1 

(c) Imbalance (in some cases refer to non–uniformity)

P  P 
U  dB  I.Lmax  I.Lmin  10 log  3  for bar state and 10log  2  for cross state
 P2   P3 

 P   
   P    
 Crosstalk: CT  dB   10log  3 1  for bar state and 10log  2 1  for cross state
 P1  2  
   P1  2  

In easy terms, CT levels generated by a single wavelength input at the undesired output port

can be given as [5].

 P   
   P    
CT  dB   10log  3 1  for bar state and 10log  2 1  for cross state
 P1  1  
   P1  1  

There are other types of losses that are incurred during switching and may influence switch
performance, like reflectance or return loss, directivity, isolation factor etc. Other parameters
that can also be taken into account include scalability, reliability, energy (power) usage and
temperature resistance [1]. The term scalability refers to the ability to integrate several smaller
(1×2, 2×2) switches into a large non- blocking (i.e. no possible routing path is blocked) N×N
switch. It is particularly an important concern. Reliability is another important issue, which
relates the lifetime operation (generally 20 years) of the switches within acceptable
performance. The solid-state technology without moving parts allows almost unlimited
switching states without any wear and tear [37]. The losses discussed above do not represent
all sort of losses incurred during optical switching, as many losses also take place within the
switch due to signal attenuation, scattering, improper design and fabrication losses. Material
dependent absorption losses and improper doping level in the base material also makes a
contribution in degrading its performance.
All optical switches

All optical switches


2.1 Introduction

A switch takes traffic from an input port or connection and directs it, over a fabric, to an output
port. Existing electronic switches handle variable-length packets, fixed-length cells, and
synchronous time slots to perform these operations. An optical switch, on the other hand,
works with light and used to direct a single wavelength, or perhaps a range of wavelengths,
from input port to output port. The all-optical switch is a more recent development. This is an
all-analog device, where both the input/output modules and the backplane are optical. The
primary benefit of all-optical devices may be their greater scalability over electronic switches.
In fact, all-optical switches are completely bit-rate and protocol transparent. We have to
realize, that most of the technologies of the all-optical switches are still emerging and usually
exist in a sub-optimal form with use of amplification rather than regeneration to boost signals
during the switching. All optical switches can be utilized to disconnect, bypass and reroute
fiber optic communications. This chapter gives introductory detail for different types of all
optical switches and their principle of operations, followed by a brief discussion on popular
architectures/structures proposed in past to build higher order switching matrix using an
elementary switch.

2.2 Geometries of all optical (O–O–O) switches: Types and applications


A general and broad classification of optical switches categorizes them into mechanical type
and non-mechanical type. In mechanical switches, switching can be achieved with the help of
moving optic fibers or optic elements using mechanical or electromagnetic means. In non
mechanical switch categories, the more important ones are electro-optic switches, thermo-optic
switches, acousto-optic switches and semiconductor optical amplifier (SOA) based switches.
Switches based on moving fibers tend to be relatively slow and are hence only useful for
routing of an optical transmission path, such as routing around a fault, while switches based on
electro–optic (EO), thermo–optic (TO), magneto–optic (MO) and SOA–switches may be used
to perform fast switching and are best suited for modern network applications and logic
operations. In this section, we have explained various structures, which are useful to design all
optical switches with their operating principles.

2.2.1 Opto-mechanical Switches

In opto-mechanical switches the switching process is made by mechanically moving miniature


mirrors, prisms, lenses etc. in and out of the optical path or by using directional coupler. It is
one of the earliest techniques and not very popular now. These switches are fabricated on
silicon with the use of MEMS technology [40]. In these switches, microscopic mirrors are
arranged in a crossbar configuration as shown in figure 2.1. When a mirror is activated, it
moves into the path of the beam and directs or diverts the light to one of the output paths/ports.

Figure 2.1 MEMs switching technology; ref. [1]


All optical switches

Mechanical switches have certain desirable features like low insertion losses, low polarization
dependent loss (PDL), low crosstalk (<–50 dB), and they are inexpensive devices [1]. These
are useful for fiber protection and wavelengths add/drop applications. The main disadvantages
are their low speed (~ 10 ms), sensitivity to vibrations, beam divergence, mechanical wear and
tear, lack of scalability and bad integrability with other optical components, that’s why, not
suitable for large switch architectures. In most cases, the switch configurations are limited to
1×2 and 2×2 port sizes, thus rendering them incapable for higher order switching requirements
of today’s emerging technologies.

2.2.2 Thermo-optic Switches

These switches are essentially constructed on waveguide material whose refractive index is a
function of the temperature. The main principle of working is the change of refractive index of
the dielectric material with the change in the temperature of the material itself. These switches
are generally small in size with high-driving-power characteristics and require forced air
cooling for reliable operation. TO–switches can be classified into two basic types:
Interferometric and Digital optical switches, shown by figure 2.2 (a) and (b) respectively. In
interferometric switches, relative phase difference is achieved by varying the refractive index
in one arm of the interferometer, which is then used to switch an input signal from one port to
another. Digital optical switch is consisting of two interacting waveguide arms through which
light propagates and the phase error between the two arms determines the output port [1].
When the temperature of one arm is changed it induces a change in its refractive index and the
light goes from one path rather than the other. An electrode arrangement is used to generate the
phase error due to EO or TO induce index variation in one arm, which guides the beam to the
output port.

(a) (b)
Figure 2.2 Thermo-optic (a) Interferometric switch (b) Digital optical switch, ref [1].

Other kinds of thermo optic switch are thermo-capillarity optical switches, thermally generated
bubble-type switches, and TO–switches with coated microsphere resonators [40]. Main
advantages of theses switches include their polarization-insensitive operation and fast
switching speed, which is of the order of a millisecond (usually < 5ms) [1], [37], [39].
However, their limited integration density and high-power dissipation make them little
impractical for on–chip system applications. Besides, air cooling is required for electrode
operation as they have typically high driving power characteristics.

2.2.3 Electro-optic Switches

Electro–optic properties such as refractive index of certain materials can be changed by


modulating the electric field. This produces a bending effect on light transmitted through such
material and in turn can be used for switching. For example, the coupling ratio of a directional
coupler based switch can be changed by varying the refractive index of the material in the
coupling region. The change in the refractive index guides the light through the appropriate
All optical switches

waveguide path to the desired port. Some of the popular switching structures based on electro-
optic effect are depicted in figure 2.3 (a–d). In all of them, the switching response time is
typically less than 1 ns, making them suitable for optical packet switching [5], [22], [41]. Other
switches based on EO–effects include liquid crystal (LC) switches, electro-holographic optical
switches, semiconductor based and electronically switchable waveguide Bragg gratings
switches etc. [40]. However the use of this property to build large networks for routing of
signals requires large arrays of switches. Using a myriad of crystals is not suitable, that’s why
LC–switches are preferred only where fast switching operation is not required such as optical
add–drop multiplexing in optical fiber networks. An effective solution to the problem is
integrated optics technology. Integrated optics waveguide switches are fabricated using EO
dielectric substrates, such as Lithium Niobate (LN) with the waveguides represented by strips
of relatively higher refractive index generated by diffusing titanium or proton exchange onto
the substrate [1], [3–4], [15–18], [27–29].

(a) (b)

(c) (d)
Figure 2.3 Electro-optic switches (a) Interferometric switch (b) Directional coupler based switch (c) Y-branch
(DOS) switch (d) SOA-based switch; ref. [5]

Various structures such as Mach-zehnder interferometer (MZI), Multimode interferometer


(MMI), directional couplers, etc. have been employed along with the application of this
principle to fabricate optical switches with many advantages in switching speed, E.R. and
device size, etc. However few of the problems with these switches such as high insertion loss
and high crosstalk need to be taken care of for efficient operation [42]. Another application of
EO-effect is in the form of a digital optical switch, which works on mode evolution and gives
steps-like response to a controlled electric field [17–18], [27–34], [43]. SOA based switches
are also termed as current-controlled optical switches, in these few SOAs are used to turned
OFF–ON by the bias current controlled methods [9], [44]. However, these are expensive
components and vulnerable to polarization losses due to the highly directional orientation of
the laser active region. Besides, it also introduces additional complexity in the system.

2.2.4 Other switching technologies

Acousto-optic (AO) switches operate on the principle of Bragg deflection of light by sound. In
these switches, the sound intensity controls the power of the deflected light, as in the switches
based on TeO2 crystals. The angle of deflection is controlled by the frequency of the sound
[45]. In these switches, input light signal is divided into its two polarized components (TE and
TM) and directed to two distinct parallel waveguides. By virtue of an AO–effect in the
material, this leads to the formation of an equivalent of a moving grating. At a selected
wavelength, this moving grating can be phase-matched to an optical wave [1]. Phase matched
signal is the lower and nonmatched is the upper output as shown in figure 2.4 (a). The
switching speed of acousto-optic switches is limited by the speed of sound and is in the order
of µs. This presents a very significant difficulty in its use in high speed optical networks.
All optical switches

(a) (b)
Figure 2.4 (a) Acousto-optic switch; ref. [1] (b) Plasmonic switch; ref. [49]

Recently, very small and extremely fast optical switches based on surface plasmons (SPs) are
reported at [46-49]. The term surface plasmons describe coherent electron oscillations at the
interface between any two materials (e.g. a metal-dielectric interface, such as a metal sheet in
air) [49]. When these electron oscillations are coupled with photonic energy, this result into a
hybridized excitation, termed as surface plasmon polariton (SPP). This SPP can propagate
along the surface of a metal until energy is lost either via absorption in the metal or radiation
into free-space [49]. An SPP waveguide all-optical switch is shown in figure 2.4 (b). In this, a
layer of photocromatic molecules (PCs) is provided to interface the path optically for control
pump. In the transparent state (ON state), the SPPs freely propagate through the molecular
layer, while in the absorbing state (OFF state), the SPPs are strongly attenuated [49]. In this
thesis, we have designed and charaterized the optical switches based on EO-effects on the
material chosen.

2.3 Higher order switch architectures

Various architectures have been proposed to design higher order switches (larger than 2×2).
All of these architectures consist of appropriately cascading small switches into a predefined
interconnecting pattern like crossbar, Benes, Spanke, Spanke-Benes, Banyan etc. [1]. The main
constraints while building large switching blocks are their blocking characteristics, switching
time and loss uniformity and the number of basic switch elements and crossovers required
[50]. We shall be discussing these constraints in details before explaining some popular
architecture available to design higher order switches.

Constraints for higher order switch design


For ultrafast communication networks as wavelength division multiplexing (WDM), dense –
WDM (DWDM), etc., switches with large port counts are required to perform desired
switching with a fast response. However, there are many constraints presenting a tradeoff
while designing the optical switches with high port count. The following constraints are mainly
suggested by G. I. Papadimitriou et al. in Optical Switching published in 2007 [1].

Number of small switches required


Large optical switches have been popularly made by cascading low port count switches such as
2×2 or 1×2 switches. Therefore the cost of these, to some extent is proportional to the number
of such switches required. The number of small switches can vary widely across architectures.
This is the primary factors that affect the system cost. Other factors that also influence the
system cost are packaging, splicing, and ease of fabrication.

Loss uniformity
A significant problem in addition to losses is the possibility that switches may have different
losses for different combinations of input and output ports. The situation takes a far graver turn
All optical switches

for large switches. The minimum and maximum number of switch elements in the optical path
for any combination of input and output ports usually offers a pretty good measure of the loss
uniformity and it is highly desirable that this number should be nearly constant. The loss
uniformity becomes more significant for higher order switches (N×N).

Number of crossovers
A common practice for fabricating large optical switches is to integrate multiple switch
elements on a single substrate. There is a fundamental difference between fabrication processes
in optical switches and integrated circuit (IC) technology, where connections between the
various components can be made at multiple layers, whereas in integrated optics, all these
connections need to be made in a single layer by means of waveguides [1]. The overlapping of
paths of two waveguides introduces two undesirable effects: power loss and crosstalk. It is thus
desirable to minimize, or completely eliminate, such waveguide crossovers although the
possibility of eliminating crossovers might be heavily dependent on switch architecture.

Blocking characteristics
Switches can be broadly classified into two major classes on the basis of the switching function
achievable: blocking or nonblocking in the presence of other lightpaths. The defining criterion
for a nonblocking switch is if an unused input port can be connected to any unused output port
irrespective of the light-path currently configured. Thus, for a nonblocking switch there is no
interconnection pattern between the inputs and the outputs that is not routable, while in a
blocking switch there will be at least one interconnection pattern that cannot realized
physically. Most applications require nonblocking switches. If rerouting of existing
connections is needed to achieve the nonblocking property, then the switch is known as
rearrangeably nonblocking switch. These switches require less number of small switches to
build a larger switch of a given size. Certain applications usually have no difficulty in using
architectures with rerouting of connections, while others may have stringent requirements.
Since existing connections must be interrupted, at least briefly, in order to switch it to a
different path, it presents an increased complexity in routing algorithm and also momentarily
disrupts traffic flow. Since optical switches are not very large, the increased complexity may
be acceptable, depending on circumstances [1]. However many time critical and sensitive
applications will not allow existing connections to be disrupted, even temporarily, to
accommodate a new connection. Usually, a tradeoff has to be considered between these
different aspects. The most popular architectures for building large switches [50] are the
crossbar.

Connecting patterns for higher order switch design


Using integrated-optic technology switches can be scaled up on a single-substrate. Scaling up
of optical switches is however limited by two factors i.e. the size of the basic switching
element and on-chip planar nature of the interconnections. These constraints can be overcome
at least partially by using intersecting waveguides instead of parallel waveguides. The
terminals of waveguides are attached to optical fibers. The rectangular structures inherent in
the technology used to fabricate large optic switches makes the coupling of optical fibers
inefficient. At this point insertion losses are observed. The polarization also plays an important
role here as the coupling coefficient is dependent on it, which makes it absolutely essential that
states of polarizations are preserved at the connecting terminals. Usually higher order switches
(more than two in/out ports) are designed by interconnecting small switches using a predefined
connecting pattern, which are also responsible for the cost of the entire network. The selection
of connecting patterns strongly depends upon the applications where the switch is needed to be
incorporated and the flexibility of expansions for future use. There are many types of
All optical switches

connection patterns, which has been used for design and fabrication of higher order switching
matrix, some of them are mainly suggested by G. I. Papadimitriou et al. in Optical Switching
published in 2007 [1].

Crossbar Connection

In this type of architecture, the routing between the input and the output ports is achieved by
appropriately setting the states of the 2×2 switches that form a unique connection for each
input-output combination. Figure 2.5 (a) depicts a 4×4 crossbar switch, which requires 16, 2×2
switches to achieve all I/P and O/P interconnection patterns. For example, to connect input i to
output j, the path traverses the 2×2 switches in row i till it reaches column j and then traverses
the switches in column j till it reaches output j [50]. One of the advantages of crossbar
architecture is that it is wide-sense nonblocking [1] and can be fabricated without any
crossovers. For crossbar architecture, the shortest path length is 1 and the longest path length is
2N–1, making this variation significantly high with increasing values of N. Due to this
significant variation in path lengths for different combinations, switching times and losses for
different switching routes will be different. This architecture has potentially the most non
uniformity in switching time and losses, limiting its use in most cases.

(a) (b)
Figure 2.5 (a) Crossbar connection (b) Benes architecture, ref. [1].

Benes connection
The Benes network is the lowest-cost switching fabric using N×2logN, known to yield
operation without any internal blocking [51]. Figure 2.5 (b) depicts an 8×8 switch, realized
with 20, 2×2 switches using Benes architecture. This is the most efficient connecting pattern in
terms of the number of 2×2 switches it uses to build larger switches [49]. An N×N Benes
switch requires (N/2) (2log2N–1) 2×2 switches, being a power of 2. The architecture is
rearrangeably non blocking switch architecture and also exhibits better loss uniformity. The
loss is the same through every path in the switch; each path goes through (2log2N–1) 2×2
switches. Its two main drawbacks are that it is not wide-sense nonblocking and that a number
of waveguide crossovers are required, making it difficult to for integrated optics applications
[1].

Spanke connection
Spanke architecture has been gaining popularity for designing higher order non-integrated
switches. This architecture is strict sense nonblocking and requires 2N (N–1) 1×2 switches,
and each path has length 2log2N as shown in figure 2.6 (a–b). Depending upon the number of
splitters and combiners, it can be classified as symmetric or asymmetric type. A symmetric
type nonblocking N×N switch requires equal number (N) of 1×N splitters and N×1 Combiners
[52].
All optical switches

(a) (b)
Figure 2.6 (a) An Spanke switch (N×N) and (b) 1×8 switch using 2×2 SEs; ref. [1] [49]

Spanke–Benes connection
Spanke-Benes architecture incorporates the properties of both the crossbar and Benes switch
architectures. It is rearrangeably nonblocking and requires N(N–1)/2 switches with no
crossovers as shown in figure 2.7(a), which represent an 8×8 switch using 28, 2×2 switches.
The shortest path length is N/2, and the longest path length is N. Its main drawbacks are that it
is not wide-sense nonblocking and possess non uniform losses [1].

(a) (b)
Figure 2.7 Switches (8×8) with (a) Spanke-Benes (b) Banyan connecting patterns, ref [1]

2.3.2.5 Banyan connection


Banyan connecting pattern as shown in figure 2.7 (b) is blocking in nature, as there is only one
path between each input and output port. But as paths can be encoded as a string of
consecutive output labels of switching elements (SEs) [1], they act as self-routing switches
(rearrangeably nonblocking) with each SE that accepts a packet using the destination address
in the packet header to identify the output port for sending the packet [1]. We have
summarized all discussed architectures in terms of important selection issues within table 2.1.
Table 2.1: Comparison of architectures used to design higher order switches, ref. [1]
Architecture Name Non blocking type No. of 2×2 SE Longest path Shortest path
Crossbar Wide sense N2 2N–1 1
Spanke Strict sense 2N(N–1) 2 log2N 2 log2N
Benes Rearrangeable N(2log2N-1)/2 2 log2N–1 2 log2N–1
Spanke-Benes Rearrangeable N(N–1)/2 N N/2
Banyan Rearrangeable (N/2) × log2N log2N log2N
Materials and their switching effects

3.1 Introduction

Selection of materials for the purpose of switching operations in present fiber-optic


communication systems which operate in the near-infrared region with low attenuation optical
windows (850nm, 1310nm, 1550nm and 1625nm) has become an important issue to avoid
possible losses during switching and for their compatibility with the present fabrication
technologies. Materials which can refract, reflect, transmit, disperse, polarize, detect, and
transform electromagnetic radiation in ultra violet, visible or infrared spectral regions are
termed as optical materials. They are used for fabricating fibers, small dimensional waveguides
and optical elements such as lenses, mirrors, windows, prisms, polarizers, detectors, and
modulators.

Their optical properties like refractive index, transparency, spectral dependency, uniformity,
strength, hardness, temperature limits, chemical resistivity etc. are determined by microscopic
level investigation of interaction between atoms, their electronic configurations and photons.
These properties can be altered or controlled by varying the wavelength of the incident light,
other parameters like temperature, pressure and in some cases by applying external electric or
magnetic fields on the material. This chapter deals with prominent characteristics of optical
materials which makes them suitable for all optical switching. Glass, crystalline materials,
polymers, plastic materials, composite semiconductors, synthetic organic crystals etc. are most
commonly used materials for fabricating optical devices and elements. A variety of plastic
materials has been used for fabricating economical and light but uniquely designed optical
elements showing high precision. But they are susceptible to microscopic defects (result in
light scattering), stresses (birefringence) and temperature variations (change in the refractive
index). Many compound semiconductors such as GaAs, GaAlAs, and InGaAsP, etc. have been
used to fabricate lasers, light emitting diodes, and photodetectors. Compositions and
architectures, which are not possible with inorganic materials, synthetic organic polymers such
as lithium fluoride, calcium fluoride, and potassium bromide, alkali-halide crystals, etc. have
replaced natural crystals to fabricate durable, optically efficient, reliable, and inexpensive
photonic and optoelectronic devices. Recently, birefringence crystals have also been used for
fabricating splitters, modulators, switches etc. These crystals such as NH4H2PO4 (ADP),
KH2PO4 (KDP), KTiOPO4 (KTP), β-BaB2O4 (BBO), Li2B4O7, LiNbO3, LiTaO3, etc. behave
nonlinearly, when exposed to very strong field.

3.2 Switching phenomena in optical materials


Switching operation takes place inside the switching fabric on application of an optical signal.
This signal is provided by the control module. In case of line switching, signal reacts to
signaling information, whereas in optical packet switching signal reacts to control information
contained in the header. In each case a reconfiguration of the optical path has to be
implemented through the fabric switch at that particular input. A switch can be classified by
three main aspects [53]:
 The switching mechanism,
 The material and
 The switching structure.

To explain the classification, note that the refractive index of a waveguide changes e.g. on
application of mechanical, thermal or electro-optical effects, thus, changing the optical path
length and causing signal phase change, which can be exploited by switching structures or a
pair of coupled waveguides such as MZI, MMI, Y-junction waveguides, etc. The used material
will determine the amplitude of the control signal needed to obtain the required refractive
index change, as well as for influencing the signal quality and the switching time. Thermo-
optic control of optical waveguide devices is an attractive procedure for the sake of simplicity
and flexibility offered by it. The following principle in general used for fabricating TO–effect
based optical devices.

3.2.1 Thermo-optic effect

The thermo-optic (TO) effect is the thermal modulation of the refractive index of a material
[54], [55] and is present in all practically used waveguide materials. In TO materials, the
refractive index is a function of temperature and can be represent simply as.

n (T) = n + α. T........................ (3.1)


dn
where   is TO-coefficient. The material index (n) changes with temperature (T) due to
dT
the change in the density (ρ) and due to the temperature change itself. This effect is mainly
used in interferometric configurations because of their architecture, which ensure sufficient
space between heaters so that they do not influence neighboring switches when implemented in
materials with relatively high thermal conductivity such as silica. The switching speed depends
upon the rate of heating. As polymers have a relatively large coefficient of expansion, i.e.
strong TO–effect is exhibited and require switching power 100 times less than silica, polymer
based switches are faster than silica based switches [60].

3.2.2 Electro-optic effect

In some pristine materials, when electric field is applied, distortion of positions, shapes and
orientations of the molecules of the substance takes place due to which their optical properties
are altered. Along with other properties, refractive index of the material also undergoes change
and this phenomenon is referred to as EO-effect [1] [5] [61]. Although the change in refractive
index is relatively small but it has significant consequences on the wave travelling through it.
As EO–effect mainly concerns with the interaction of electromagnetic and the electrical
(electronic) states of materials, significant variations in absorption constant or refractive index
can be observed by applying electric field. The term comprises of a number of distinct
phenomena, which can be subdivided into:

(a) Change of the absorption:


 Electroabsorption: It is defined as the general change of absorption constants due to
application of electric field/voltage/current.
 Franz-Keldysh effect: It leads to the change in the absorption shown in some bulk
semiconductors.
 Quantum-confined Stark effect: It effects the change in the absorption in some
semiconductor quantum wells.
 Electro-chromatic effect: It is defined as the creation of an absorption band at some
wavelengths, which gives rise to a change in color.
If we choose to use a less strict definition of the EO–effect and also consider electric fields
oscillating at optical frequencies, we can include nonlinear absorption (absorption depends on
the light intensity) to category (a) and the electro-optic Pockels and Kerr effect (refractive
index depends on the light intensity) to category (b). Along with the photo–electric effect and
photoconductivity, the electro-optic effect gives rise to the photorefractive effect.

(b) Change of the refractive index:

 Linear EO–effect (Pockels cell): It describes the changes in refractive index in


proportion to the applied electric field. Solid crystals lacking inversion symmetry
exhibit the Pockels effect.
 Quadratic EO–effect (Kerr cell): It states that a change in refractive index is
proportional to the square of the applied electric field. Most electrooptic materials show
this effect.
 Electro-gyration: It is related with changes in the optical activity of the materials.

In EO–directional coupler, there is a periodic exchange of optical powers (P1 and P2) between
the two parallel waveguides during the coupling action due to field induced refractive index
variation. The strength of the coupling operation is determined by two major parameters [39]:

 The coupling coefficient ρ, which in turn depends upon the dimensions, wavelength
and refractive indices,
2n
 The difference in the propagation constants   1 –  2 
0
where Δn is the index variation. In case of identical waveguides Δβ= 0 and P2 (0) =0, then at a
distance L0 = π/2 ρ, called the transfer distance or coupling length, the power is transferred
completely from one waveguide to other.
Mach-Zehnder Interferometer based switches
4.1 Introduction
Among the various available topologies, Mach–Zehnder interferometer (MZI) based structures
are one of the most efficient to convert a phase modulation into an intensity modulation. These
structures are commonly used to build a variety of applications such as optical modulators,
splitters, switches, etc. by using electro-optic (EO) or thermo-optic (TO) effects.Utilizing the
EO-effects, the symmetric MZI structure has been preferred to design flexible and high speed
switches with low driving power requirements for smaller bandwidth operation [108–111].
Due to large EO–coefficients, lithium niobate (LN) is a suitable material for MZI structure
based switches. These devices possess stable performance parameters, even for inputs having a
wide range of optical power levels [112]. These switches use metal electrodes over the
interferometric arms that generate a refractive index gradient of an optical medium under the
influence of an electric field, resulting into a bending effect on light transmitted through the
medium, termed as phase modulation. In this chapter, design and analysis of an EO–MZI
switch has been done using theoretical approach of MZI structures and rigorous design
modeling steps. The switch performance is enhanced by optimizing various design parameters
such as indiffusion process parameters, waveguide dimensions, applied field etc. The switch
performance has been evaluated for its operations for low attenuation optical windows. In the
next part, we have explained the effect of tapered (asymmetric) interferometric arms and
crystal cut on the MZI switch performance. These are also useful for integrated optic
applications to realize large switching matrix. To explain this, we have recalled the
interconnected switching fabrics to design completely nonblocking, reconfigurable 8×8
Banyan switch using MZI as an elementary switch. In the later part, we have discussed a
SOA–MZI compact switch [9] by exploiting the band gap shifting character of SOA.

4.2 Switching in MZI structure

The MZI switch can operate in the bar state, where most of the light appears in the output
waveguide on the same side as the input, and the cross state, where most of the light moves to
the output waveguide on the other side. The switching in a conventional MZI structure can be
achieved by introducing a path delay among the propagating beams either by inserting a delay
element as show in figure 4.1 (a) or with TO/EO effects induced index change in the
interferometric arms.

4.2.1 MZI switch with path delay element

Wavelength-dependent switches can be made using either active or passive MZI structures. In
passive MZI structures, phase modulation is achieved by inserting a delay element or changing
the optical path length directly. While in an active MZI structure phase or intensity modulation
is achieved with an external field source such as EO or TO–effects to generate delay in
propagating length in a indirect way. Figures 4.1 (a–b) illustrate the constituents of a 2×2
passive MZI structure. This consists of three stages: an initial 3–dB directional coupler which
splits the input signals, a central section where one of the waveguides is longer by L to give a
wavelength-dependent phase shift (delay) between the two arms, and another 3dB coupler
which recombines the signals at the output. The recombined signals interfere constructively at
one output and destructively at the other if the delay is chosen properly to deliver the optical
power only at one output port. In the end of interferometric region, the outputs from its two
2 neff
arms have a phase difference of   k.L , where k  . Therefore by producing a delay

in the propagating path, signal can be switched from one port to another port.

(a) (b)
Figure 4.1 Layout of (a) MZI structure and (b) Path delay with unequal interferometric arms.

4.2.2 Electro-optic MZI Switch

A conventional EO–effect based 2×2 MZI–switch (passive MZI structure) consist of two
interferometric arms of equal length connected between two 3dB couplers as shown in figure
4.2. These arms are spaced wide enough so that there is no evanescent coupling between them.

Figure 4.2 EO-MZI switch with equal interferometric arm lengths, ref. [1]

Its operation can be understood as follows: the first coupler is used to divide the light evenly in
two parts, which when passed through the interferometric arms experience a net phase change
of 2 . This phase difference is due to a push–pull effect caused by the field applied in
opposite directions through the waveguides under the electrodes [79]. As it has been observed
with this structure that, the output intensity is periodic with minima and maxima occurring at
odd and even integer multiples of applied voltage. Therefore the light can be received at output
ports by constructive or destructive interference phenomena [1]. Accumulation of a phase
difference between the two arms causes the recombined light to interfere according to the
following equation [113].
Pout 1+cos( )
 …………….. (4.1)
Pin 2

Various materials have been used with their EO properties to design and implement MZI
structure based optical switches for two most popular optical windows i.e. 1.3 µm and 1.55
µm.

Calculation of the power imbalance of first stage 3-dB coupler

In absence of modulating electric field, all proposed MZI switches with equal interferometric
arms remain in cross state. Therefore, first we will evaluate the impact of the power imbalance
of first stage 3–dB coupler on the CT levels, which are generated at the outer facet of the
interferometric arms. An MZI structure obtained by splicing of commercial fused couplers
suffers less isolation factor, which is a difference between optical losses of the two arms often
occurs from the splices [100]. Reduced isolation ratio leads to a reduction of the optical signal
to noise ratio (OSNR), which increases power levels of destructive interferences at the MZI
outputs. This will further reduce power level difference of constructive and destructive
interference of the MZI, leading to an increased bit error rate (BER) at the reception. A defect
in coupler coupling ratio (ideally50/50, 3dB) also affect the quality of the interferometer [100].
Similarly, mismatching of couplers due to variations in the coupling region (narrowing the
waveguides at coupling region), result in degradations of switch performance. Also, if the
input optical field amplitude is not split exactly equally between the two arms, as is often the
case, the minimum power out will be greater than zero, and the switch will not be completely
off. The extinction ratio is then not infinite, but can easily made to be greater than 10 to 20 dB
[79].

Figure 4.8 In and Output definitions of both couplers (3–dB splitter and 3–dB combiner)

As shown in figure 4.8, we have first evaluated coupler (3–dB splitter) performance in terms of
power level difference at the output (P3 and P4), which is termed as imbalance that result in to
different power levels (Q1 and Q2) at the input ports of second coupler, i.e. at the end of
interferometric arms of the structure. The imbalance factor (non-uniformity) and the generated
CT levels due to Q1 and Q2 are calculated as follows [5].

 Imbalance at the output of 1st 3-dB splitter

P 
U  10log10  3  ……… (4.7)
 P4 
 The CT levels at the end of interferometric arms, prior to 2nd 3-dB combiner.
2
 Q1 1/2
 Q 
1/2

   – 2 
CTINF = 10log  2   2 
 ……. (4.8)
10
1/2 2
 Q1 1/2
 Q 2  
     
 2   2  

where Q1 and Q2 are power levels at the end of interferometric arms and are calculated as

follows [5].

1 ………………… (4.9)
Q1 
10 (U/10)
 1

Q2  1– Q1 ……………………….. (4.10)
Multimode Interference based switches
5.1 Introduction

With advent of self–image formation [136–138] in multimode waveguides, multimode


interference (MMI) structures have attracted considerable interest due to their unique
characteristics such as compactness, relaxed fabrication tolerance, large optical bandwidth, and
polarization insensitivity [10], [139−142]. Therefore, they have been preferred for designing
many optical components like optical couplers, splitters, multiplexers, optical gates, switches,
etc. [143−146]. Advancements in MMI−structure have made it possible to design ultra−short
couplers and switches with segmented, cascaded and integrated structures [66], [147]. MMI
based couplers are often used to replace traditional 3−dB splitters and combiners in MZI–
switches [67], [148] to make them less polarization dependent and tolerant to fabrication
errors. Principally, the MMI couplers can be used as a switch, by eliminating optical
confinement to change the waveguide behaviors or by altering the imaging length by changes
in refractive index [149]. These structures are emerging as a preferred choice to design reliable
and polarization insensitive mixers, splitters, switches with low crosstalk levels than their
counterparts like directional couplers, Y−shaped waveguiding structures etc. due to ease of
their fabrication and integrability with optical networks [150]. Similarly, tapering techniques
within conventional MMI structures would lead to device designs with reduced beat length,
better loss uniformity and improved E.L. [151−153], which is relatively unique with respect to
other structures. This chapter deals with MMI structures and use of their self−imaging
principle to design basic 2×2 MMI−switches by defining the partial refractive index regions in
the coupling area. The channel waveguides are optimized to reduce overall switch area
coverage and to enhance switching performance. Initially, the switches are simulated for a
range of wavelengths, centered at 1.55µm and then the structures are further modified to
operate with other test wavelengths. These switches are then further used within a defined
connecting pattern to implement and analyze higher order switch matrices. In the end, we have
proposed 4×4 switch designs using cascaded SEs to check their uniform loss response, with
low onchip area coverage. Though, cascading of MMI single SEs leads to a large device size,
higher losses and less robustness to fabrication errors.

MMI−switches

The use of light pipes to generate multiple self–images of symmetric objects was first
suggested by Bryngdahl [136]. Ulrich et al. [137–138] was able to extend the principle to the
replication of images of random objects in multimode waveguides and fiber interferometers
based on analogous phenomena. Based on this phenomenon, they have also demonstrated a
3−dB directional coupler and a wavelength separator (filter) and possible crossover of strip
guides. However, towards early 1990s only, the concept of self–imaging in uniform index slab
waveguides has been studied in more details. Soldano, Penning, et al. [139−141] explained its
mathematics by calculating coupling coefficients and locating multiple images inside the
waveguides, which is further used in [154−155], to elaborate the phase relation between the
input, the outputs and transition inside the MMI region.
5.2.1 Concept of self images in MMI structures

The centre structure of an M×N MMI device is a waveguide designed to support a large
number of modes (typically≥ 3) [139]. There are M single – mode input waveguides that are
connected to a central structure which is connected to N equally distanced single mode
waveguides spaced. This central structure is known as coupling region and its length is called
as coupling length. The light enters into the central structure and retrieved from it through the
input and output waveguides respectively. Generally, self imaging is the principle behind the
operation of MMI−switches. According to this principle, along the direction of propagation of
the multimode waveguide, an input field profile is reproduced in single or multiple images at
regular intervals. Following theory for formation of self images in multimode waveguides
explained here is mainly suggested by L. B. Soldano et al. reported in 1992 and 1995 [10]
[139].

Figure 5.1 Field pattern in 2×2 MMI–coupler, ref. [10]

Figure 5.1 shows the field pattern of a 2×2 MMI−coupler. In this device, the width of the step
index multimode waveguide is WM and the refractive indices of core and cladding region are nr
and nc respectively. It can accommodate maximum m lateral modes at a wavelength λo, which
can be characterized by mode numbers v = 0, 1… (m − 1). The relation between the core index
nr, the propagation constant Bv and the lateral wave number k, can be given by the expression
below [10].
k xv2  v2  k02 nr2 …………. (5.1)

2 (v  1)
where k0  , k xv 
0 WM
Assuming that the input field profile ψ(x, z) is confined within WM , includes guided modes

and radiative modes then at z=0 (fig.5.1) it can be expressed as:

 ( x, 0)   (cv v ( x)) ....... (5.2)


v

where ψv is the normalized distribution of vth mode’s modal field. Again, using overlap

integrals, we can express the field excitation coefficient, cv as below [10] [139].

cv 
 ( x, 0)( ( x))dx …... (5.3)
v

 ( x)dx
2
v

At a distance z = L, the field profile is estimated as [10]


 v(v  2) 
 ( x, L)   v 0 ( v ( x)cv ) exp  j
m 1
L  ……… (5.4)
 3L 
where v denotes mode number and Lπ denotes beat length for two lowest−order modes. Further,

the determination of the shape of ψ(y,L), and the types of images formed can be done by the

field excitation coefficient cv and the mode phase factor

 v(v  2) 
exp  j L  ………. (5.5)
 3L 
Therefore, we can change the field profile by modulating mode phases and obtain the desired
output field. For certain conditions, the field profile at Ψ(x, L) is a replica of the given input
field at x=0 (Ψ(x, 0)). As explained by Ulrich and Soldano et al. [137−139], the field profile at
the input (Ψ(x, 0)) is replicated at regular intervals producing single and multiple mirror
images according to the following conditions. As shown in Fig. 5.1. (2×2) 3−dB couplers can
be realized using this phenomenon of two−fold imaging. The use of these single and multiple
images in 3db, bar and cross couplers can be summed up as following.

 When L = 2p (3Lπ), the field reproduced at L is in phase with the field at the entrance, and the
coupler is in bar state.
 When L = (2p + 1) (3Lπ), the field reproduced at L is in antiphase with the input field, and the
coupler is in cross state.
 When L = (p + 1/2) (3Lπ), the field is linear combination of input field and its mirror image in
x−z plane hence coupler is in 3db state.

5.2.2 Tunable MMI−switches

As explained earlier, MMI structures usually realize optical functions by directly tuning the
refractive index of the index modulation (IM) region, located within the MMI section. The
switching mechanism in most cases of MMI structures is the same, i.e. they operate by
modifying the refractive index at specific areas within the MMI waveguide, which are
collocated with the occurrence of multiple self−images. This change in the refractive index is
the prime factor behind altering the phase relation between the self−images, which ultimately
modifies the output image and switches the light between the output waveguides. A. M. Al–
hetar et al. [142] has nicely explained, the classification of MMI–switches in accordance with
IM locations, which is as follows:

In size modulated MMI–switches, the IM regions are located horizontal to the direction of
propagation of light. Changes in the refractive indices of the segments (IM regions) inside the
MMI section also bring about variation in the width of the MMI region. In first configuration
shown by figure 5.2 (a), the confinement guide region is created to allow the light to pass
through the region. The use of different interference phenomena in a device allows achieving
different switching states. If the width of the MMI regions is reduced by depressing the
refractive index (by means of the EO/TO–effects) then the imaging locations will be changed.
Though, EO–effect is highly effective to use in this type of MMI–switch in comparison with
TO–effect., in which case the power consumption is also high due to the length of index
modulation. This type of switch offers compactness and multifunctionality and is hence very
promising for future DWDM and OXC systems.
(a) (b)
Fig. 5.2 (a) Size modulated (b) Image modulated 2×2 MMI−switch, ref. [8] [142].

In MMI structures, the output images are resulted due to formation of new self images in
subsequent intervals which originates from the interference pattern of the self–images in first
interval. Consequently, modification of the refractive index around some selected spots within
one interval of the MMI, where such self–images occur, can be used for getting desired output
images [142]. Based on this approach, a 2×2 image modulated MMI–switch is shown in figure
5.2 (b), in which the IM region is located where two images are formed. For waveguide length
equal to odd multiples of Lπ, the phase shift between symmetric and asymmetric modes is an
odd multiple of π, and the input image will be inverted. Introducing an additional phase shift of
π, at one of the two formed images will make the image switch to the other port. However, this
approach works properly only when the refractive index change is entirely confined within the
areas containing the principal of self–imaging [114].
Digital optical switches
6.1 Introduction

Since its invention, the digital optical Switch (DOS) has proved to be a very attractive
component for space switching and multi−wavelength optical communication systems
[171−173]. It has certain favorable characteristics such as step like switching response, low
polarization and low wavelength sensitivity, making it highly insensitive to wavelength,
polarization, refractive index, temperature and other physical parameters that may affect
switching. As DOSs functioning is based on mode evolution, therefore high drive voltage (or
power) is required to operate these switches as compared to interference based switches.

6.2 Splitter and switches with Y−junction waveguides:

A conventional Y−junction guiding structure usually consists of tapered waveguide sections


connecting access input waveguide and two output waveguides, as depicted in figure 6.1. In
this waveguiding structure, spacing between its outer waveguides increases linearly. At the
junction point, equal amount of light is launched into each waveguide. With no applied
voltage, the branch is geometrically symmetric and acts as an equal power splitter [31], evenly
distributing the optical input power into the two output ports.

Figure 6.1 Layout of a conventional Y−junction EO−DOS

On application of proper biasing across the electrodes, one of its branches becomes electrically
asymmetric due to the EO (or other) effect, and the light is guided adiabatically to a branch
which has a higher refractive index [17]. The guidance arising due to this mode evolution will
be even more effective if it is possible to further increase the refractive index of the channel
waveguide, relative to the other port. This sort of ideal behavior will take place as long as the
geometric transition represented by the branch is sufficiently adiabatic to prevent the mode
coupling between the normal modes [31]. On reversing the asymmetry, the light can be guided
to the other branch. Thus the biasing across the respective branch plays a key role in the
switching action. For single mode branch arms, the switching due to mode evolution is not
expected to be strongly dependent on wavelength or polarization, except through the usual
dependency of the EO−coefficients [43]. Figure 6.2 depicts a typical Y−junction waveguides
based compact optical power splitter. It exhibits a better response for a wide range of applied
optical power. To a certain extent, it is independent of wavelength and less sensitive to
polarization and surrounding temperature.

Figure 6.2 Optical power splitter with Y−junction waveguides


To facilitate the comparison of our designed switches presented in this thesis with those
previously reported, we shall be evaluating splitter performance with following standard
definitions.
P 
Imbalance  dB  10log10  1  ……………… (6.1)
 P2 
P  P  
Splitter efficiency  %    1 2  x100 .….. (6.2)
 P0 
P0
Ideally, P1  P2  , i.e. the splitter delivers half power exactly at both output ports, therefore,
2
the power imbalance and efficiency shall be 0dB and 100% respectively. In the waveguide
based power splitters, their response can be optimized by adjusting the coupling function and
waveguide shape in a way that results in high split efficiency and minimum power imbalance
at the output ports. However, optical losses take place during the splitting of the input power at
junction point and also during their delivery to destination output ports. Practically, these
losses can be kept at minimum levels by either having a control over the design and diffusion
process, or by providing a supplement to the losses.

You might also like