|
In the past several years the growth
of optical networks and the amount of data being transmitted
through them has generated great interest in new means of
managing network traffic. One of the most promising is the
optical cross-connect (OXC), which is designed to enable the
switching of light signals from a group of input fibers to
a group of output fibers with no transition to electrical
signals in between, as is the current standard.
It is believed that avoiding the transitions to and from
electrical signals will offer a number of advantages. This
all-optical method is bit-rate and protocol transparent, which
means the OXC is able to function even as the bit-rate grows
and as the data transmission protocol changes. In addition,
direct optical switching is highly scalable and provides telecommunication
carriers with a value proposition that avoids the expensive
high-speed electronics used in current optical-electrical-optical
(OEO) topologies.
There are a number of technologies that provide for direct
optical switching including liquid crystals, bubbles, holograms,
and microelectromechanical systems (MEMS). Of these approaches,
MEMS is widely believed to be the most promising for large-scale
optical cross-connects. 
Figure 1: 2D cross-connect in two different states.
MEMS for Optical Cross-Connect Switches
The two major approaches for MEMS OXCs are the so-called 2-D
and 3-D approaches. The 2-D approach involves arranging a
set of MEMS mirrors in a plane. Each mirror is placed at a
fixed angle with respect to the incoming light and is moved
in and out of the path of the light like a shutter. By cascading
sets of these mirrors one is able to steer the input from
M input fibers to N output fibers. For the case in which M=N,
the total number of mirrors required is equal to N-squared.
The so-called non-blocking Clos architecture can reduce the
number of mirrors but adds a large amount of complexity to
the system. A diagram of a simple 3x3 switch is shown
in Figure 1 with a total of 32=9 mirrors.
Two different states are shown. In the first
state, each input fiber is connected
to its companion output fiber. In the second state, the
outputs of fibers 1 and 3 are switched. This approach works
quite well for small port counts (N<32) but as the port number
grows, the system requires many more mirrors and the total
path length increases, which increases the insertion loss.
At these higher port counts, 3D MEMS technology is the preferred
choice. 3D MEMS arrays operate by steering beams of light
in an analog fashion in 3-dimensional space. For N fibers
in and N fibers out the total number of mirrors needed is
2N and the length that the light travels does not increase
as quickly with port count as it does for 2D configurations.
The general configuration for such an array
is represented in the schematic of Figure 2, which shows an
arrangement with 3 mirrors on each array. The diagram shows
a standard arrangement of the arrays with respect to the fibers
and how the mirrors rotate to switch the output fiber to which
each input fiber is connected. Each mirror can rotate about
2 axes and can direct incident light to any mirror in the
companion array. The mirror in this array then redirects the
light so that it enters the correct output fiber at normal
incidence. The fundamental functionality required of each
mirror thus is the ability to rotate to many different positions
in both the x and y-axes and hold that position for extended
periods of time (up to years) with high accuracy. The fact
that the mirror is functioning in an analog fashion greatly
increases the complexity of the system.
The mirrors for these arrays typically consist of silicon
plates coated with a film that has a high reflectivity at
the wavelengths used in the network. Each mirror is suspended
in space by springs that are compliant enough to allow the
mirror to rotate in response to forces applied to it. Because
the mirror must be able to rotate about two axes, it is often
suspended within a gimbal, which is itself suspended from
 a
fixed structure as shown in Figure 3. This structure allows
the mirror to rotate about one axis within the gimbal, which
in turn can rotate about the perpendicular axis. In this way
the mirror can achieve compound angles of rotation.
The means by which the actuation of the mirror is achieved
can be of many forms but often falls into one of two categories:
electrostatic and electromagnetic.
Figure 2: 3D cross-connect in two different
states.
Electrostatic Actuation
Electrostatic actuation consists of applying a bias between
two conductors, which induces an attractive force between
them. Whether achieved through a parallel plate or comb-drive
architecture, this actuation scheme is very commonly used
and has a number of advantages including low power consumption
and relative ease of design. The parallel plate architecture
is more straightforward and easy to fabricate but suffers
from high-voltage requirements and strong non-linearity of
force versus displacement. The comb-drive architecture is
more linear and requires smaller voltages, but it is significantly
more difficult to fabricate.
Electromagnetic Actuation
The other main actuation type is electromagnetic, which uses
the interaction between an electromagnet and a permanent magnet
to rotate the mirror. These devices have the advantages of
relatively large torques when used with larger mirrors, and
a smaller number of leads per mirror. They do, however, consume
more power, which can lead to challenges of heat dissipation.
They can also be more complicated to package, as they typically
will involve the assembly of external magnets or coils.
Critical Figures of
Merit for 3D MEMS:
There are many requirements imposed on the 3D MEMS that form
the core of an OXC. Some of the most important of these are:
Maximum angle- the most basic requirement for each
mirror in an array is that it can rotate enough to direct
light to any mirror in the opposing array.
Mirror size/fill factor- the size of the mirrors in
an array must be large enough to capture a large percentage
of the incoming light and is thus heavily dependent on the
optical design of the system. For a given mirror size, it
is often desired to have a high "fill factor" which
is the area of each mirror divided by the area of each pixel.
ROC (Radius of Curvature)- in order for the mirrors
to be sufficiently reflective, they are coated with a film,
which is typically gold. This film will induce some curvature
on the mirror, which can be quantified by its radius of curvature.
The curvature can induce an unwanted spreading of the reflected
light. Therefore, it is necessary to keep this curvature low,
i.e. keep the ROC high.
Switching speed- to fit into the protocol of many optical
fabrics, the switching speed often must be less than 10 ms
to ensure uninterrupted service. To achieve such switching
speed, MEMS designers reduce the mass of the rotating structures
while increasing the maximum torque applied to them.
Pointing stability- as the mirror drifts from its ideal
angle, the light it deflects is steered away from the center
of the fiber to which it is directed causing a decrease in
signal strength. The extent to which angular error manifests
itself in an increase in insertion loss is critically dependent
on the optical design of the system. This requirement can
be quite challenging under shock, vibration, and temperature
cycling and is one of the main drivers for closed loop control.
Scalability- one of the advantages of 3D MEMS is its ability
to scale to higher port counts without any radical change
in its implementation. However, a number of challenges do
arise at higher port counts ranging from increased angular
deflection and pointing stability requirements to the difficulty
of routing an increasing number of leads.
Reliability- this area is very broad and is influenced
by almost every aspect of system design. Reliability tests
are governed by the conventional Telcordia specifications.
These are used to qualify a system to ensure that it meets
standard telecommunication reliability requirements. MEMS
designers have the challenging task of meeting these explicit
requirements and any additional ones proposed by a specific
customer. This often involves a careful choice of materials
whose properties change little over time, the use of small
actuation signals (whether voltage or current), and hermetic
packaging of the device to minimize environmental effects.
Figure 3:
SEM micrograph of micromirror array.
Summary
MEMS offer many advantages as a technology platform for optical
switching and specifically as the core of a large port count
OXC. As with any emerging technology, the 3D MEMS OXC will
ultimately succeed or fail based on its ability to meet customer
demands for functionality and price. The success in meeting
these requirements will have a great impact on the field of
optical networking as well as on the future of MEMS itself.
Dr. Thomas Kudrle is the Lead Engineer of MEMS OXC at
Corning IntelliSense Corporation. For more information contact
Dr. Kudrle at KudrleTD@corning.com
or by telephone at (978) 988-8000, ext. 2311. Visit IntelliSense
at www.intellisense.com or contact the main office at IntelliSense
Corporation, 36 Jonspin Road, Wilmington, MA, 01887, USA.
|
