Ezekiel Kruglick
Two- and three-dimensional MEMS arrays form the building blocks of free-space optical switching modules. A mastery of MEMS design, packaging, and volume production skills is essential to produce cost-effective, reliable devices.
Optical crossconnects based on microelectromechanical systems (MEMS) use arrays of tiny mirrors that move in either one or two dimensions, and are designated a two-dimensional (2-D) or three-dimensional (3-D) design, respectively. The newest wave of switching components are micromachined mirrors, fabricated on silicon chips. A 2-D design can be used for optical crossconnects containing up to 128 x 128 ports, while the 3-D option is more scalable and preferable for crossconnects with several thousand ports.
MEMS-based all-optical switches are based on the principle that a moving mirror routes photons from one optical fiber to another. The routing is accomplished by steering the light through a collimating lens, reflecting it off a movable mirror and directing the light back into the output ports.
CHOICE OF MATERIALSMicroelectromechanical systems exploit well-established, low-cost silicon very-large-scale integration foundry processes, which have been used for years by manufacturers to provide high volumes of low cost, reliable processors. Leveraging the billions of man-hours invested by the integrated-circuit (IC) industry, MEMS technology has emerged as the clear choice for designers when building a scalable optical-switch fabric.Most of the current crop of MEMS devices use silicon as the primary mechanical material, a mature material proven by decades of use and research as well as years of applications. Polycrystalline silicon (polysilicon), which has been used for years in producing the gates of transistors, and single-crystal silicon (the material of which wafers are made) are the two practical types of silicon for most MEMS applications.
In general, polysilicon is easier to design and manufacture. Polysilicon is so trusted that many automotive airbag systems rely on polysilicon sensor structures. It is stronger than steel, shows few if any long-term aging effects and is repeatably manufacturable.
2-D DESIGNS
A t the heart of the 2-D switch is an array of digital micromirrors. Each mirror in the "on" state directs a collimated beam from an input fiber to an output fiber. The mirror array is controlled by a simple digital interface and provides strictly nonblocking, low-loss connections. Because there is no electronic conversion of the optical signal, the switches can accommodate any data rate or format to meet rapidly changing network requirements (see Fig. 1).
As shown in Figure 1, the fiber connections are controlled through an array of micromirrors (diagonal lines) on a microchip (center) that are selectively activated to connect the desired fiber. All signal paths are bidirectional. Note that the switching function only occurs between planes 1 and 2. The 2-D approach offers a very simple interface. Micromirrors and fibers are arranged in a planar fashion and the mirrors can only be in two positions: either on or off.
An N x N array of MEMS micromirrors is used to connect N input fibers to N output fibers. This is called N-squared architecture, as it uses N2 individual mirrors. The advantage of this approach is that it requires only simple controls consisting of transistor-transistor-logic (TTL) drivers and associated electronics to provide the required actuation at each MEMS micromirror.
Because the MEMS mirrors are so light, they can be switched in milliseconds, offering unparalleled full-fiber switching speeds. The general rule is the lower the mass, the higher the speed. This speed potential provides another motivation for using relatively thin polysilicon instead of single-crystal silicon.
Two-dimensional designs also support the introduction of a third and fourth fiber port to a basic N x N switch. This configuration permits dynamic add/drop functionality and arrays of 1 x N switches in a single package. The multiplane configuration also allows 2-D switches to be combined or cascaded to form large-scale switch fabrics.
3-D DESIGNSWhile the 2-D approach is simple and inexpensive in low port counts, 3-D scales well into very large port counts (thousands of ports). The 3-D approach uses the same principle of moving a mirror to redirect light, however the mirror is not constrained to two positions but can redirect light to any position in two dimensions (see Fig. 2). This results in 2N architecture, because two arrays of N mirrors each are used to connect N input to N output fibers. For example, a 256 x 256 switch would only require 512 mirrors.The 3-D method is less constrained than the 2-D method by the scaling distance of light propagation as the port count grows. Such architecture can scale to thousands by thousands of ports with low loss and high uniformity.
At the core of the 3-D switch is an array of scanning micromirrors. One option is to base each mirror on a double-gimbaled design, which allows high-resolution switch-path positioning from any input to any output, in a manufacturable and reliable cell.
ACTUATION METHODS
Possible methods for actuating a MEMS optical switch are electrostatic and magnetic. Electrostatic actuation, based on the attraction of oppositely charged mechanical elements, is currently the best method of actuation. Commonly used for actuating MEMS devices, its advantages include repeatability, ease of shielding, and well-researched and understood behavior. This method of actuation is built into huge numbers of high-reliability devices each day and has demonstrated long lifetimes and fundamental reliability.
Magnetic actuation relies on attraction between magnetic materials and typically one or more electromagnets. This method can generate larger forces with high linearity. However, issues of poor shielding, crosstalk fabrication, and unproven reliability have kept magnetic MEMS actuation out of most proven applications so far. One other point to note is that the use of magnetic actuation can strongly impact power expenditure because several amps can be required to drive each mirror—a consideration that rapidly becomes serious as systems require thousands of mirrors.
Finally, of course, the only way to convince users that a device can provide reliability is through long-term testing and field use. So far, electrostatic is the primary mass-produced and fielded MEMS actuation method.
QUALITY CONTROL AND RELIABILITY
With both 2-D and 3-D designs, subsystem manufacturers face tough challenges with respect to packaging and automation if delivery of switching products in high volume will be possible. The importance of this challenge cannot be overstated. Designing MEMS for mass production can be a very different process than designing a single MEMS device.
For people accustomed to designing for the IC world, where scaling up production is relatively easy, the transition to MEMS is more difficult. Adopting a mindset on mass production from the beginning of development is a necessity.
Optoelectronic packaging continues to remain a challenge for designers. The small size of MEMS and the immense numbers of devices inhabiting a single chip make them perfectly suited to optical switching, but also create the greatest difficulties in packaging.
MEMS structures are sensitive to humidity and dust. To protect against these potentially destructive agents, MEMS chips must be protected by a hermetic package seal. Optical-MEMS switching modules and subsystems must have a hermetic design that allows multiplane optical-fiber arrays to interface with the package—raising the design issue of achieving seals between fibers that seal the material and packaging to guarantee high reliability and yield.
Additional points of weakness are introduced by the need for multiple fiber and electrical feedthroughs in the hermetic package. Each fiber or pin entering or leaving the unit can present a compromise to hermeticity. Careful attention must be paid to this issue in the manufacturing process and it can determine the necessity of integrating electronics.
AUTOMATION IN MANUFACTURING
It's easy to let manual manufacturing steps inflate cost and slow production. Whatever the device, poorly designed MEMS can require many hours of assembly to make them ready for operation. Each part requires many operator hours, which could potentially balloon costs and reduce yield.
Automation is the key to high-volume, low-cost manufacturing. It dramatically reduces production time and handling, while ensuring reliability, precision, process consistency, and quality. Each of these factors brings valuable gains.
The semiconductor industry realized the necessity for automation many years ago and the MEMS industry must follow suit. Wherever possible, the workforce should avoid actual manufacturing but instead be overseeing the various processes including troubleshooting and quality control. Automation presents the key to controlling manufacturing processes and ensuring hermeticity while producing the high volumes required for current and future applications.
The opportunity for the design and manufacture of MEMS devices for all-optical switches is very great and immediate. Manufacturers who succeed in driving the development of optically switched networks will demonstrate an in-depth understanding of MEMS design, as well as the packaging and production of qualified and reliable devices in high volume. In the atmosphere of fierce competition, early and careful consideration of high-volume production will make the designer's job simpler and much more effective.
Ezekiel Kruglick is the principal MEMS designer at OMM, 9410 Carrol Park Drive, San Diego, CA 92121. He can be reached at 858-362-2800.