Access Official Technical PDF

Share Summary:LinkedInEmail

An all-optical XOR gate can compare two optical data streams without first converting them into electrical signals. In a semiconductor optical amplifier-based Mach–Zehnder interferometer (SOA-MZI), the inputs alter the phase of optical probe pulses, and interference turns those changes into a logical output. Numerical analysis of this architecture examined how device and signal parameters affect performance at 20 and 40 Gb/s, offering practical guidance for designing high-speed optical logic.

How an SOA-MZI produces an XOR output

The gate uses a symmetrical Mach–Zehnder interferometer, with one semiconductor optical amplifier in each arm. The two data inputs, A and B, act as control signals. A separate clock pulse stream serves as the probe and is split between the arms. At the output, the probe signals recombine and interfere.

When A and B have the same logical value, the two arms are intended to remain balanced, so the signal at the selected output port is suppressed. When the inputs differ, one arm experiences a different gain and phase response from the other. This difference arises principally from cross-phase modulation (XPM) in the SOAs. The resulting phase offset changes the interference condition and allows probe light to emerge as the XOR output.

The analysis considered two arrangements. In a traveling-pulse MZI (TMZI), control and probe pulses travel in the same direction through the SOAs. In a colliding-pulse MZI (CMZI), they travel in opposite directions. The CMZI configuration was a particular focus because, in the described arrangement, a filter is not required at the output to reject the control signals.

Why phase, gain and timing matter

An SOA’s response is dynamic: an incoming control pulse changes the carrier population, affecting both optical gain and phase. The simulations modeled these effects while accounting for SOA length and nonlinear gain compression. They assumed identical SOAs, synchronized inputs and maintained polarization alignment. Loss, amplified spontaneous emission noise, group-velocity dispersion and polarization-dependent effects were excluded, so the results describe the modeled conditions rather than every practical implementation.

The key performance measure was output contrast ratio (CR), defined as the minimum output peak power for a logical one divided by the maximum peak power for a logical zero, expressed in decibels. A higher contrast ratio indicates clearer separation between the two output states. The study used this measure to compare parameter choices under pseudorandom input data patterns.

SOA-MZI design considerations for an all-optical XOR gate

SOA length and carrier lifetime

Shorter SOA length generally improved the simulated gate performance. Length had a stronger effect in the CMZI than in the TMZI: in the counter-propagating arrangement, the gain conditions vary along the amplifier as the pulses pass one another. For the modeled 20 Gb/s case, an SOA length below 250 μm and a carrier lifetime below 150 ps supported a contrast ratio above 8 dB under the specified conditions.

Carrier lifetime also affects how quickly the amplifier recovers between pulses. A longer recovery time can broaden or slow the switching window, increasing the risk that a response to one bit interferes with the next. At 40 Gb/s, the parameter limits become more restrictive in the simulations; the reported design guidance includes a carrier lifetime below 100 ps.

Linewidth enhancement, confinement and gain

The linewidth enhancement factor, commonly denoted by α, links changes in carrier density to changes in optical phase. Increasing α can produce a stronger phase shift, but the relationship is not simply “more is better.” In the 40 Gb/s simulations, values above about 4 distorted and broadened the switching window, contributing to trailing-edge intersymbol interference (ISI).

The optical confinement factor and small-signal gain also influenced contrast and switching-window shape. Under the tested conditions, lower confinement factor and higher small-signal gain favored better performance. The study’s 40 Gb/s parameter guidance included a confinement factor below 0.35 and small-signal gain above 25 dB. These values are simulation-specific design indications, not universal component specifications.

Pulse width and control-pulse energy

Shorter control and probe pulses generally supported faster operation. For the modeled 40 Gb/s case, pulse widths below 6 ps were among the suggested conditions. However, extremely short pulses were not automatically beneficial: when pulse width fell below 2 ps, the simulations showed degradation associated with a secondary switching window.

Increasing control-pulse energy improved performance only up to a point. Beyond a certain level, further increases produced little additional benefit in the modeled results. The findings also indicate that the differential phase shift does not always need to reach an ideal odd multiple of π for useful switching. This matters because it suggests that satisfactory contrast can be achieved without requiring the maximum possible phase excursion.

Pattern dependence and the 40 Gb/s challenge

SOA gain does not recover instantly. Runs of consecutive one bits can drive an amplifier further into saturation, while zero-bit intervals allow gain to recover. As a result, the phase difference between interferometer arms depends partly on recent data history—not just on the current input pair. The simulated output therefore showed pattern-dependent peak power, including residual output during some zero states when both inputs were identical.

This behavior can close the switching window incompletely and create intersymbol interference. The simulations found that larger carrier lifetime and smaller small-signal gain could slow the window’s falling edge, while a larger linewidth enhancement factor could produce an additional switching window. Such effects help explain why a design that performs adequately at 20 Gb/s may not maintain the same output quality at 40 Gb/s.

Practical implications for optical logic design

The analysis indicates that an SOA-MZI can support simulated XOR operation at 40 Gb/s when its optical and amplifier parameters are selected together. No single value determines performance: SOA length, carrier dynamics, gain, confinement, pulse width and input energy interact. The results also depend on assumptions such as synchronized signals, identical amplifiers and the exclusion of noise and certain nonlinear effects.

For engineers evaluating optical logic, the practical lesson is to optimize the full switching response—not phase shift alone—and test it against realistic data patterns. Contrast ratio and window shape both matter, particularly as bit rates rise and gain recovery becomes more consequential. The reported parameter ranges provide a starting point for SOA-MZI design analysis, while implementation decisions require validation against the characteristics and operating conditions of the actual optical components.

Access Official Technical PDF

About The Author