Optical router operates at 250 Gbits/sec

By YVONNE CARTS-POWELL

Optoelectronic switching nodes currently used in fiber-optic communications networks are a bottleneck to system throughput. A completely optical packet-switching method that operates at 250 Gbits/sec has been developed by professor Paul Prucnal and researchers Ivan Glesk and Koo Il Kang at Princeton University, Princeton, NJ. In a paper in the electronic journal Optics Express (paper 2225, published by the Optical Society of America, at www.osa.org), the researchers explain how the system achieves all-optical address recognition using a self-routing scheme.

Two address bits from each packet header are used for routing. The two packets are composed of bits spaced 4 psec apart. The optical routing switch drops photonic packets from the network traffic at their destinations. The packet-switching bit-error rate has been measured at less than 10-9.

Although the group had previously developed methods for ultrafast optical switching using time-division multiplexing, they had not demonstrated optical control of the process.

Why optical?

Both packet- and circuit-switched communications networks now use electronic nodes connected by optical links. Each electronic channel, however, has a limited bandwidth of a few tens of gigahertz, so most switching architectures use high dimensionality (many components working in parallel) to provide high throughput.

If the technology is developed, all-optical nodes could provide high throughput with fewer switches, allowing use of the full bandwidth available in optical fibers. Such systems would also remove the optoelectronic interfaces and electronics needed for routing controls, thus decreasing the hardware complexity and failure rate of the node. This requires optical technology that can switch packets, process signals, and recognize the address of the packets, thus allowing control of the switches.

Optical switching technology, in the form of lithium-niobate cross-bar switches, is fairly well developed. (Unlike electronic switches, it has proven easier to build optical switches with high throughput than with high dimensionality.) In the system demonstrated, these devices are switched using electrical pulses.

Optical-signal processing associated with switching avoids the data flow bottleneck at the input to the photonic switch and reduces the need for flow control. The researchers report, "We have demonstrated ultrafast optical flow control using a recently developed Terahert¥Optical Asymmetric Demultiplexer," also called toad. The device can demultiplex a picosecond time-slot from a nanosecond address-frame. It requires less than one picojoule of switching energy and can be made small enough to be integrated onto a chip.

In ultrahigh-speed optical networks, address bits in a compressed packet are spaced only picoseconds apart. To read the address bits, ultrafast demultiplexers (such as the toad) are used, one to read each address bit in the packet header. This address information is used by the routing controller to set the switch.

The researchers demonstrated the optical routing using a network of 2 ¥ 2 lithium-niobate cross-bar switches, an ultrafast all-optical address recognition unit (which decodes the destination address), a routing controller (which sets the state of the switching element), and an optical buffer that matches the delay of the input packets to the processing delays of the routing controller.

Polarization is key

To test the system, 1-psec pulses at 1.313 micron are generated from a 100-MHz-repetition-rate neodymium-doped yttrium lithium fluoride laser. Each pulse is divided into four pulses: three packet bits and a clock pulse. Loop polarizers set the clock and data pulses into orthogonal polarization states, and a polarization combiner is then used to multiplex the clock and the packet header data onto a common transmission line.

"This is the self-routing feature of the system," says Kang. Because the position of the clock pulse signals the location of the address bits in the packet, changes in the length of the transmission line (which can change the amount of polarization rotation in the line) are not critical.

Polarization splitters are used to identify the address bits and forward each one to a toad to be demultiplexed. The toad outputs are detected using photodetectors and a fast and gate. The demultiplexed address bits trigger a pulse generator, which in turn sets the switch to a cross or bar state.

Crosstalk measurements of pseudo-random data in adjacent 4-psec-width time slots of the packet header, equivalent to routing errors, exhibit a bit-error rate of less than 10-9, with strong jitter immunity. q

Yvonne Carts-Powell writes on photonics issues from Belmont, MA.