SPIE Membership Get updates from SPIE Newsroom
  • Newsroom Home
  • Astronomy
  • Biomedical Optics & Medical Imaging
  • Defense & Security
  • Electronic Imaging & Signal Processing
  • Illumination & Displays
  • Lasers & Sources
  • Micro/Nano Lithography
  • Nanotechnology
  • Optical Design & Engineering
  • Optoelectronics & Communications
  • Remote Sensing
  • Sensing & Measurement
  • Solar & Alternative Energy
  • Sign up for Newsroom E-Alerts
  • Information for:
SPIE Photonics West 2019 | Register Today

SPIE Defense + Commercial Sensing 2019 | Call for Papers

2019 SPIE Optics + Photonics | Call for Papers



Print PageEmail Page

Optoelectronics & Communications

High-Speed Optical Switches

From OE Reports Number 187 - July 1999
30 July 1999, SPIE Newsroom. DOI: 10.1117/2.6199907.0001

Figure 1. (A) Correlator chip (0.797 cm X 1.71 cm) containing two FLC-VLSI SLMs and a CMOS camera. (B) Miniature correlator optical system including a correlator chip with attached diffractive Fourier transform lenses (bottom) and two mirrors with attached polarizers (top). The penny illustrates the small size of the system.

The Internet, multimedia, telecommuting, and image processing applications are driving the need for bandwidth in displays, communications, and computing beyond the levels that even the most optimistic scientists would have believed only five years ago. Growing at 300 percent per year, the Internet will redefine how we share data, conversations, and images. It will literally reweave the fabric of our lives into completely new shapes and patterns.

Out of the display, communications, and computing trilogy, electrical engineering and computers stand most ready to meet the challenges proposed at the end of the communications line. The need for an all-optical computer has basically gone the way of the Hula-Hoop; however, the data lines that run between workstations are quickly becoming overloaded with data, surpassing the ability of strictly electronic switching capabilities.

For instance, a standard electronic switch in a communications device can achieve 50-GHz switching speeds. However, telecommunication companies are already deploying the first terabit-per-second optical network. Within the next few years, experts anticipate bandwidth of telecommunications systems -- already beleaguered by high-cost electronic switches, routers, add/drop multiplexers, and demultiplexers -- will exceed the capability of electronic switches.

The answer lies in an all-optical network where light packets self-route themselves along trunk lines to their local area destination.

Switching crystals

Jay Damask of Lucent Technologies Inc. (Murray Hill, NJ) explained that a typical telecommunications system could make use of several levels of switching technology, from silicon microelectromechanical (MEMS) devices to mechanically tuned fiber Bragg gratings and semiconductor devices. Alternatively, ferroelectric liquid display technology may make inroads to switching fabrics.

While all these methods show some level of promise and MEMS technology is the emerging leader, Damask said that as yet it is unclear what the winning technology will be.

High-speed ferroelectric liquid crystal on very-large-scale integration (FLC-on-VLSI) devices with 100-msec. switching speeds are already finding applications in high-definition television sets, said Mike O'Callaghan of Displaytech Inc. (Longmont, CO). Unlike their nematic counterparts, FLCs can have an electric dipole moment, which means they can be driven from "on" to "off" very quickly by changing the polarity of the electric field applied to the molecules. Nematic liquid crystals do not have an electric dipole moment and are instead driven by making use of their anisotropic dielectric constant, which means they can be turned "on" very quickly by an electric field of either polarity, but then slowly relax to an "off" state when the electric field is removed.

FLC-on-VLSI technology may be the one way in which optics contributes to computing technology through the development of optical processors (Figure 1a and 1b). Displaytech sells FLC-on-VLSI displays with 1280x1024-pixels that could be used as spatial light modulators that run as fast as 3000 frame per second, requiring 3.9 Gb/s of data. A standard 33-Mhz PCI bus can supply data to a processor at rates of only 1.06 Gb/s; therefore, optical correlators based on such SLMs would already be ahead of electronic technology and could add considerable processing capabilities to image-processing applications where many feature extraction's or filter operations need to be completed very quickly.

Displaytech's FLC-on-VLSI devices are already being used by Samsung Electronics to develop new rear-projection high-definition TV sets through the use of special algorithms that take the raw binary speed of the FLC device and convert it to 24-bit color images at normal video rates using field-sequential color and duty cycle modulation.

Fiber switches in telecom

Researchers at the Univ. of Michigan, in conjunction with corporate partners, are working on fiber switches that use Sagnac interferometers in a nonlinear optical-loop mirror configuration that delivers switching speeds in excess of 50 Gb/s (Figure 2).

Mohammed Islam and his associates in Ann Arbor, MI, are working on several all-optical device configurations to meet the needs of the next generation Internet and communications technologies. From continuum laser sources that emit across the telecommunications window (from 1.3 to 1.6 um) and Raman amplifiers that compliment the industry standard erbium-doped fiber amplifiers, the group is ambitious. However, they are tempered with realistic goals and input from industry.

"In the communication field, [devices] have to work in the teleco window between 1.3 and 1.6 µm," Islam said. "It has to be compatible with that because no one is going to wavelength switch for the purpose of a switch. It has to be fiber based or fiber in, fiber out...and it definitely has to be [wavelength-division multiplexed] compatible.

"The fourth element is it has to be inherently serial in nature because fiber is serial in nature," Islam said, "and it has to be commensurate with the bit rates going down the fiber. At the very minimum, that's tens of gigabits."

In addition to performance, Islam said timing for a viable all-optical switch is crucial. "If what you're proposing is material science effort, it just doesn't compute because of the time scale," he said. Islam estimates that all-optical switches need to be ready to go in the ground by 2005 when Internet applications will likely drive commercial communications systems upwards of 10 Tb/s. "So, basically, you've got fiber optics or three-five semiconductors," he said. "Silicon germanium doesn't work in the wavelength window."

According to Islam, these kinds of optical switches make use of the nonlinearity that arise from the third-order dielectric susceptibility, where the frequencies remain the same for both the input and output and the nonlinearity refers to a change in the materials index of refraction as a function of intensity.

The nonlinear optical mirror switch under development at the University of Michigan is basically a Sagnac interferometer that uses a local control pulse to control the path of the input signal. This device consists of a four-port direction coupler, where a loop of fiber is coupled to two ports: one for the input signal and another for the output signal. When acting as three-part switch the input pulse enters the loop traveling in the same direction as a control pulse from a local source. The signal pulse, orthogonal in polarity or frequency, phase shifts the control pulse that travels in the same direction because of the nonlinear index of refraction. A portion of the control pulse is then extracted by a polarized beam splitter and travels on its way. In the absence of a control pulse, however, the signal pulse continues to travel along a preselected course.

Optical loop mirrors based on the Sagnac interferometer, as first proposed and demonstrated by researchers in the Optics and Quantum Electronics Laboratory at MIT, can also be used as optical regenerators since the high-intensity pulse represents the data and low-intensity dispersive pulses are the noise. Because only high-intensity pulses are transmitted, the noise is filtered out, leaving the original optical signal to be amplified by either an erbium-doped fiber amplifier or Raman amplifier, depending on the wavelength in question. Electronic signal regenerators placed every 600 km or so represent a significant bottleneck and cost for WDM telecommunications applications.

According to Islam, similar nonlinear optical configurations can be used as Boolean switches for dimultiplexing and local area routing when the pulse is a soliton. Called a soliton-dragging NOR gate, two separate fibers (one for the signal and the other for control) are coupled, ending in a polarizer (Figure 3). The signal pulse is orthogonally polarized to the control pulse. If the control pulse reaches the coupler at the same time as the signal pulse, cross-phase modulation causes a shift in the central frequency of the signal pulse, changing the speed of the pulse through the fiber because of soliton-group-velocity dispersion in the fiber. If the frequency -- and therefore speed-is changed, the pulse does not arrive at the destination within a given digital clock window, and therefore represents a 0. If the control signal does not affect the input signal, then the pulse arrives at the proper moment giving a "1" value.

Semiconductor switches

Counter to the all-fiber approach is the use of semiconductor waveguides as nonlinear directional switches. Based on tried-and-true manufacturing processes, AlGaAs and GaAs/AlGaAs offer bandgaps that can be tailored to the telco wavelength window. The nonlinear directional coupler is one such switch. In design, two parallel waveguides are created in the semiconductor material. When the light passing through the two waveguides (control and signal) is of low intensity, the two pulses couple through overlapping evanescent fields. However, when one of the pulses is of high-intensity, the nonlinear index blocks the coupling of the two pulses.

Semiconductor router switches such as this have demonstrated quick response to intensity changes in the 100-femtosecond realm, which is associated with changes in the electron cloud based on intensity variations. According to Islam, using solitons has also been shown to be important in these devices, because Gaussian pulses will switch differently from the center to the tail, leading to pulse break up, poor contrast ratios, and noncascadable (series of multiple) switches. Solitons switch as a unit, however, eliminating this problem.

"These days, you can't look at the switch for the sake of the switch," Islam said. " You must look at the implications of it, such as insertion loss. And you have to look at its impact on the overall systems, because if the switch kills your signal-to-noise ratio, companies won't look at it."

Recently, telecommunications system modeling software has become publicly available, giving high-speed optical switch designers a leg up towards successful commercial development of their components.

Contact Info:
Lucent Technologies

Bell Labs Innovations
600 Mountain Ave.
Murry Hill, NJ 07974
Phone: (1) 800/458-2368

Displaytech, Inc.

2602 Clover Basin Dr.
Longmont, CO 80503
Phone: (1) 303/772-2191
Fax: (1) 303/772-2193

The University of Michigan

Optical Sciences Laboratory
Electrical Engineering & Computer Science
1110 EECS
Ann Arbor, MI
Phone: (1) 734/647-3875

R. Winn Hardin

R. Winn Hardin is a science and technology writer based in Jacksonville, FL.