X.D. Cao

All-optical packet-drop demonstration using 100-Gb/s words by integrating fiber-based components

Packet-drop function for a time-division multiplexing network using 100-Gb/s, 8-b words is experimentally demonstrated by integrating all-optical header processing and payload demultiplexing with electrooptic packet routing. The header processor consists of two levels of all-optical logic gates based on low birefringent nonlinear optical loop mirrors (NOLMs), and the payload demultiplexer is a two-wavelength NOLM. Both devices are driven by synchronized lasers with timing jitter under 1 ps. The contrast ratios for both header processor and demultiplexer are 10:1 and that of the packet router is 17 dB. The switching energies for header processing and payload reading are 10 and 1 pJ/pulse, respectively.

Cascadability and functionality of all-optical low-birefringent nonlinear optical loop mirror: experimental demonstration

We verify both experimentally and theoretically the cascadability and the Boolean complete functionality of the low-birefringent nonlinear optical loop mirror (low-bi NOLM) operating with all inputs at the same wavelength. We achieve a low switching energy by using a low-birefringent (Δn∼3.5×10-6), polarization-maintaining fiber to achieve a longer interaction length between two orthogonally polarized pulses. We experimentally demonstrate switching in the cascaded operation of two low-bi NOLMs using picosecond pulses from an erbium-doped fiber laser. This has the potential to have a bit rate of 100 Gb/s. After the two cascaded low-bi NOLMs, the performance is a peak switching contrast of 36:1 and a timing window of 1.7 pulse widths with a switching energy of 9 pJ. In addition, we demonstrate the and and the xor/not operations with the low-bi NOLM showing Boolean completeness. The and operation has a switching contrast of 84.5:1, and the xor/not has a switching contrast of 11.5:1. Finally, we study the gate numerically and find good agreement between experiments and simulations.

Experimental cascaded operation of low-birefringence nonlinear-optical loop mirrors

We experimentally demonstrate the operation of a low-birefringence (low-bi) nonlinear-optical loop mirror (NOLM) that has the advantages of low switching energy, tolerance to timing jitter, and cascadability. Because cascading two all-optical logic gates is an important step toward high-speed optical signal processing and header processing in time-division-multiplexed networks, we also demonstrate the cascaded operation of two low-bi NOLMs. Using a passively mode-locked fiber laser that produces 450-fs pulses at a wavelength of 1.55 microm, we achieve a 10.7:1 switching contrast ratio and a 2.7-pulse-widths-wide timing window after the cascaded gates. The results agree well with theoretical predictions and confirm the advantages of the long interaction length associated with orthogonally polarized pulses in low-bi (Deltan ~ 3.0 x 10(-6)) polarization-maintaining fiber.

All-optical 40 GHz demultiplexing in a NOLM with sub-pJ switching energy

higher contrast ratios by including hard spectral apertures within the pulse shapeddecoder and to demonstrate similar thresholder concepts based on soliton self frequency shifts in fibers with A, at shorter wavelengths. We would like to acknowledge C. C. Chang for building the fiber ring laser, M. Newhouse and V. Da Silva for supplying erbium-doped and dispersion-shifted fibers, and I. Duling for helpful discussions related to erbium amplifiers. This work was supported by the National Science Foundation under grants ECS9626967 and ECS-9312256.

High-speed header processing using synchronized fiber lasers driving all-optic logic gates

We show all-optic logic gates driven by synchronized erbium-doped fiber lasers (EDFLs) to perform 100 Gb/s header processing in TDM packetized networks or optical add/drop multiplexers. We experimentally demonstrate AND and XOR logic functions in low-birefringent nonlinear optical loop mirrors (Low-Bi-NOLMs) with switching energy as low as 9 pJ and peak switching contrast of 8:1. We show that the Low-Bi-NOLMs satisfy decision-making capabilities in a header processor such as cascadability and Boolean-complete logic functions. Our novel AOM-grating scheme of synchronization provides 30 times the length and 10 times the bandwidth as compared to the scheme using piezo-electric transducers (PZTs). We have achieved synchronization of two passively mode-locked EDFLs with a timing jitter of approximately 1.3 times the pulse width. We are now integrating the synchronized EDFLs and all-optic logic gates, which are key components for packet addressing for 100 Gb/s applications.