Multi-format all-optical processing based on a large-scale , hybridly integrated photonic circuit

We investigate through numerical studies and experiments the performance of a large scale, silica-on-silicon photonic integrated circuit for multi-format regeneration and wavelength-conversion. The circuit encompasses a monolithically integrated array of four SOAs inside two parallel Mach-Zehnder structures, four delay interferometers and a large number of silica waveguides and couplers. Exploiting phase-incoherent techniques, the circuit is capable of processing OOK signals at variable bit rates, DPSK signals at 22 or 44 Gb/s and DQPSK signals at 44 Gbaud. Simulation studies reveal the wavelength-conversion potential of the circuit with enhanced regenerative capabilities for OOK and DPSK modulation formats and acceptable quality degradation for DQPSK format. Regeneration of 22 Gb/s OOK signals with amplified spontaneous emission (ASE) noise and DPSK data signals degraded with amplitude, phase and ASE noise is experimentally validated demonstrating a power penalty improvement up to 1.5 dB. ©2011 Optical Society of America OCIS codes: (130.3120) Integrated optics devices; (070.4340) Nonlinear optical signal processing. References and links 1. S. Sygletos, I. Tomkos, and J. Leuthold, “Technological challenges on the road toward transparent networking,” J. Opt. Netw. 7(4), 321–350 (2008). 2. S. J. Ben Yoo, “All-Optical Regeneration for Ultra-Long Fiber Links and its Prospects for Future Applications with New Modulation Formats,” In Optical Fiber Communication (OFC) Conference San Diego (CA), USA, March 2009, OThS4. 3. E. Kehayas, D. Tsiokos, P. Bakopoulos, D. Apostolopoulos, D. Petrantonakis, L. Stampoulidis, A. Poustie, R. McDougall, G. Maxwell, Y. Liu, S. Zhang, H. J. S. Dorren, J. Seoane, P. V. Holm-Nielsen, P. Jeppesen, and H. Avramopoulos, “40-Gb/s All-Optical Processing Systems Using Hybrid Photonic Integration Technology,” J. Lightwave Technol. 24(12), 4903–4911 (2006). 4. G. T. Kanellos, D. Petrantonakis, D. Tsiokos, P. Bakopoulos, P. Zakynthinos, N. Pleros, D. Apostolopoulos, G. Maxwell, A. Poustie, and H. Avramopoulos, “All-Optical 3R Burst-Mode Reception at 40 Gb/s Using Four Integrated MZI Switches,” J. Lightwave Technol. 25(1), 184–192 (2007). 5. V. S. Grigoryan, M. Shin, P. Devgan, J. Lasri, and P. Kumar, “SOA-based regenerative amplification of phasenoise-degraded DPSK signals: Dynamic analysis and demonstration,” J. Lightwave Technol. 24(1), 135–142 (2006). 6. I. Kang, C. Dorrer, Z. Liming, M. Dinu, M. Rasras, L. L. Buhl, S. Cabot, A. Bhardwaj, L. Xiang, M. A. Cappuzzo, L. Gomez, A. Wong-Foy, Y. F. Chen, N. K. Dutta, S. S. Patel, D. T. Neilson, C. R. Giles, A. Piccirilli, and J. Jaques, “Characterization of the Dynamical Processes in All-Optical Signal Processing Using Semiconductor Optical Amplifiers,” IEEE J. Sel. Top. Quantum Electron. 14, 758–769 (2008). 7. P. Vorreau, A. Marculescu, J. Wang, G. Bottger, B. Sartorius, C. Bornholdt, J. Slovak, M. Schlak, C. Schmidt, S. Tsadka, W. Freude, and J. Leuthold, “Cascadability and Regenerative Properties of SOA All-Optical DPSK Wavelength Converters,” IEEE Photon. Technol. Lett. 18(18), 1970–1972 (2006). 8. X. Yi, R. Yu, J. Kurumida, and S. J. Ben Yoo, “Modulation-Format-Independent Wavelength Conversion,” In Optical Fiber Communication (OFC) Conference San Diego (CA), USA, March 2009, PDPC8. #142224 $15.00 USD Received 4 Feb 2011; revised 9 Apr 2011; accepted 12 Apr 2011; published 31 May 2011 (C) 2011 OSA 6 June 2011 / Vol. 19, No. 12 / OPTICS EXPRESS 11479 9. M. Spyropoulou, M. Bougioukos, G. Giannoulis, Ch. Kouloumentas, D. Kalavrouziotis, A. Maziotis, P. Bakopoulos, R. Harmon, D. Rogers, J. Harrison, A. Poustie, G. Maxwell, and H. Avramopoulos, “Large-Scale Photonic Integrated Circuit for Multi-Format Regeneration and Wavelength Conversion,” In Optical Fiber Communication (OFC) Conference San Diego (CA), USA, March 2011, OThY2 (to be presented). 10. N. Pleros, C. Bintjas, G. T. Kanellos, K. Vlachos, H. Avramopoulos, and G. Guekos, “Recipe for Intensity Modulation Reduction in SOA-Based Interferometric Switches,” J. Lightwave Technol. 22(12), 2834–2841 (2004). 11. J. Wang, A. Maitra, W. Freude, and J. Leuthold, “Regenerative properties of interferometricall-optical DPSK wavelength converters,” Opt. Express 17(25), 22639–22658 (2009). 12. H. Kim, and P. J. Winzer, “Robustness to Laser Frequency Offset in Direct-Detection DPSK and DQPSK Systems,” J. Lightwave Technol. 21(9), 1887–1891 (2003). 13. A. Poustie, “Hybrid Integration for Advanced Photonic Devices,” In European Conference on Integrated Optics (ECIO), Eindhoven, The Netherlands, June 2008, WeB1. 14. Ch. Kouloumentas, M. Bougioukos, A. Maziotis, and H. Avramopoulos, “DPSK regeneration at 40 Gb/s and Beyond Using a Fiber-Sagnac Interferometer,” IEEE Photon. Technol. Lett. 22(16), 1187–1189 (2010). 15. R. Elschner, A. M. de Melo, C.-A. Bunge, and K. Petermann, “Noise suppression properties of an interferometerbased regenerator for differential phase-shift keying data,” Opt. Lett. 32(2), 112–114 (2007). 16. K. Tajima, “All-optical switch with switch-off time unrestricted by carrier lifetime,” Jpn. J. Appl. Phys. 32(Part 2, No. 12A), L1746–L1749 (1993).


Introduction
All-optical implementation of fundamental processing functionalities such as regeneration and wavelength conversion will be the key-factor for future optical networks of enhanced transparency, power efficiency and capability of supporting increased volume of data traffic [1,2].As optical networking is steadily evolving from static network topologies to dynamically re-configurable networks that change and adapt according to bandwidth requirements, the relevant processing subsystems should be similarly agile and reconfigurable in nature to adjust their operation under different traffic conditions in terms of data-rates and modulation formats.Furthermore, to represent a viable solution, these subsystems should be reliable, stable and easy in use, and should exhibit low power consumption and potentially small footprint.Different techniques and technologies have proved their potential for the implementation of regeneration and wavelength conversion of intensity and phase modulated signals.Nevertheless, the complete set of the aforementioned desirable characteristics clearly calls for the use of multi-functional, large-scale photonic integrated processors, based preferably on semiconductor optical amplifiers (SOAs) as nonlinear elements due to their significantly lower input power requirements today compared to their Kerr effect-based counterparts.
Simultaneous wavelength-conversion and regeneration of conventional on-off keying (OOK) signals has been demonstrated in the past using SOAs in linear or interferometric configurations.Integrated Mach-Zehnder interferometers with SOAs inside their arms (SOA-MZIs) have shown particular efficiency for 2R or 3R OOK regeneration, depending on whether these structures are fed with a continuous wave (cw) or a re-synchronized optical clock as input signal [3,4].For phase-modulated optical signals on the other hand, the processing subsystems should be able to handle both the amplitude and the phase of the signals.SOA-based regeneration of (differential) phase-shift keying ((D)PSK) signals has been demonstrated using the phase-incoherent technique that relies on decoding of the input (D)PSK signal and subsequent optical phase re-modulation within an SOA-based nonlinear interferometer [5][6][7].The specific technique may be implemented using either a Sagnac or a Mach-Zehnder interferometer, and has at least in the latter case the potential for on-chip system integration.The relevant demonstrations have been based though on discrete devices for the 1-bit delay interferometer (DI) and the SOA-MZI, as well as on bulk fibers for the interconnection of the two elements [6,7].Finally, as (differential) quadrature phase-shift keying ((D)QPSK) comes also into play, a multi-format wavelength converter relying on an optical hybrid frontend and two parallel SOA-MZIs was reported in [8].The scheme has indeed the potential for on-chip integration, yet the proof-of-principle demonstration was based on discrete devices for the optical hybrid and the IQ converter.
Recently, we made a step towards the development of agile photonic integrated processors by presenting the first silica-on-silicon chip for multi-format (OOK, DPSK, DQPSK) regeneration and wavelength-conversion [9].The operation of the multi-format processing chip (MFPC) relies on phase-incoherent techniques, thus eliminating the need for retrieving a phase-locked local carrier.Its nonlinear section consists of two parallel SOA-MZIs, whereas the linear frontend of a complex network of waveguides and DIs that drive the input signals to the SOA-MZIs after decoding.
In the present paper, we provide an analysis of the performance of the MFPC by reporting on simulations that reveal the potential for OOK and DPSK regeneration and the feasibility for efficient DQPSK wavelength conversion.Furthermore, we provide experimental validation of the regenerative performance of the device for OOK and DPSK signals at 22 Gb/s for all types of input signal distortion.Regenerative wavelength conversion with power penalty improvement of 1.5 dB for OOK and up to 1.4 dB for DPSK signals is reported.

Operating principle and device
Figure 1 presents the regeneration and wavelength-conversion techniques that are implemented by the MFPC. Figure 1a outlines the concept of OOK regeneration assuming an optical clock at λ 2 as input signal to the SOA-MZI gate.The degraded data stream at λ 1 is driven to the appropriate port of the SOA-MZI to act as control signal, and switches the pulses of the clock by means of cross-phase modulation (XPM) and cross-gain modulation (XGM) inside the SOA of the upper branch.The wavelength converted output appears at the switched port after optimum interference between the lower and upper propagating signal components at the output 3-dB coupler of the gate.Regenerative performance of degraded input signals in terms of induced amplitude-and amplified spontaneous emission (ASE-) noise is achieved due to the saturated amplitude response of the SOA-MZI that equalizes amplitude fluctuations mainly around the mean power level of the marks [4,10].
Figure 1b extends the operation to the phase-incoherent DPSK regeneration case by assuming a linear 1-bit delay interferometer (1-bit DI) that precedes the SOA-MZI gate.The 1-bit DI converts the λ 1 degraded data signal into two complementary OOK streams that are driven with their relative timing preserved to the appropriate ports of the SOA-MZI to act again as control signals.The successive pulses of the λ 2 input optical clock are switched to the upper output port and form a PSK signal, as they acquire a phase difference of 0° or 180° depending on the arm of the dominant control pulse each time.In the case that amplitudenoise is introduced to the input (D)PSK signal, the decoded streams at the output of the DI acquire intense amplitude variation that is efficiently suppressed by the subsequent SOA-MZI, by means of the same mechanism described above for the OOK signals.Although part of this input amplitude-noise induces also phase-noise variations through the SOA-MZI that is operated under saturation, overall improvement of the output signal quality is feasible due to the strong suppression of the input amplitude-noise.In the case that only phase-noise is introduced to the input (D)PSK signal, the complementary decoded OOK streams at the output of the DI exhibit smaller amplitude variations but lower extinction ratio (ER).These act as control signals in the subsequent SOA-MZI gate, which sufficiently suppresses the amplitude variations and provides an output PSK signal with well-suppressed phase-noise compared to the input signal.Finally, at the presence of ASE-noise that degrades both the amplitude and the phase characteristics of the input (D)PSK signal, regeneration potential is mainly related to the suppression of the amplitude-noise.This is in particular the case for typical SOA elements inside the MZI with moderate α-factor values [11].
Finally, Fig. 1c presents the concept of DQPSK wavelength conversion that is based on the parallel use of two linear and two nonlinear interferometers: the two 1-bit DIs operate with a relative 90° phase difference and recover the I and Q components of the input signal, whereas the complementary streams at the output ports of each DI are forwarded to the subsequent SOA-MZIs to form two PSK signals, as described above.A phase shifter allows for a 90° difference between the clock pulses entering the upper and the lower SOA-MZI enabling the formation of a QPSK signal at the final output of the circuit, from which the information of the input DQPSK signal can be recovered.As observed in Fig. 1c, the standard DQPSK decoding process results in OOK signals that exhibit a limited extinction ratio (ER) [12], having in turn as a consequence that the two PSK signals at the output of the parallel SOA-MZIs are of lower quality compared to the simpler DPSK regeneration case.However, as it will be shown through simulations in the following section, the potential for efficient wavelength-conversion is maintained.the board through a number of available input ports in the case of OOK or DPSK processing, whereas for DQPSK operation, the optical clock and the data signal should enter through port 9 or 10 and port 7 or 8, respectively.Figure 2b depicts in turn the blueprint of the 12.5x2.4cm 2 silica-on-silicon motherboard of the MFPC comprising planar silica-on-silicon waveguides and S-bend shape DIs for minimizing material birefringence.The arrow indicates the position for the placement of the monolithic SOA array by means of an intermediate silicon daughterboard [13].The SOA devices are 2.1 mm long with polarization dependent gain of about 1 dB and characteristic gain recovery time of about 20 ps with a fast time constant of less than 10 ps measured by pump-probe experiments.Finally, Fig. 2c illustrates the final packaged and pigtailed device that was used for the experimental studies presented in this paper.

Simulation results
Numerical studies were conducted to evaluate the potential for regeneration of OOK and DPSK signals and the feasibility of DQPSK wavelength-conversion using the MFPC.The assessment was based on quality (Q)-factor calculations for the OOK signals and the decoded PSK signals at the input and the output of the MFPC and additional calculations of the amplitude-and phase-noise for the DQPSK signals.The simulation setups for the different modulation formats were based on the respective diagrams of Fig. 1 with additional DI stages for evaluation of the decoded signals at the final output, whereas the SOA-model that was incorporated in the simulation platform was adjusted to emulate the experimental performance of the SOAs in terms of gain, bandwidth and recovery time.The main parameters used in the simulation model of the SOAs are summarized in Table 1. Figure 3a presents the Q-factor (20•logQ) improvement at the output of the regenerator for 22 Gb/s OOK signals degraded either with deterministic amplitude noise or with amplified spontaneous emission (ASE) noise resulting in low optical signal-to-noise ratio (OSNR) input values.Q-factor improvement in excess of 4 dB can be observed in both cases for a wide range of input values, indicating the efficiency of the technique.Figure 3b presents in a similar way the Q-factor improvement of the decoded signals at the output of the MFPC for 22 Gb/s DPSK signals suffering from variable degrees of phase-, amplitude-and ASE-noise.The highest Q-factor improvement is 4 dB, and it is observed in the case of phase-added noise in which regeneration is feasible for degraded input signals with at least 13 dB Q-factor.For lower Q-factor values, the input phase-noise induces increased amplitude fluctuations after the DI-based decoding process that cannot be sufficiently compensated by the SOA-MZI saturated amplitude response.Moreover, residual phase-noise is transferred to the output PSK signal degrading the quality of the final decoded stream [11].On the other hand, phase-noise suppression demonstrates a limited margin for improvement when the input signal is already of high quality exhibiting an input Q-factor above 25 dB.Q-factor improvement beyond 4 dB is expected with total elimination of the phase noise in the output signal by means of nonlinear elements in the arms of the MZI that are more tolerant to phase changes.Quantumdot SOAs or even quantum-well SOAs of exceptionally low α-factor values represent a good approximation to this condition [11].It is also noted that the curves for the amplitude-and ASE-noise loaded signals have a similar shape with the curve for the phase-noise loaded signals, exhibiting though a higher Q-factor regeneration threshold and lower Q-factor improvement.
Figure 4 presents the corresponding eye-diagrams in the cases of maximum Q-factor improvement with OOK and DPSK signals.More specifically, Fig. 4a and Fig. 4b illustrate the OOK signal at the input and output of the regenerator and reveal that the regenerative effect is mainly related to the suppression of the peak-power fluctuations at the mark-level.Figure 4c and Fig. 4d present in turn the decoded DPSK signals at the input and the output, and verify that the suppression of the input phase-noise appears in the form of increased ER in the decoded signal [14].Finally, Fig. 5 summarizes the results of the simulation study for DQPSK signals at 44 Gbaud.Figure 5a-5c depict the eye-and the constellation-diagram of the input DQPSK signal and the eye-diagram of the decoded I-component, respectively, for an input signal that suffers from a certain degree of amplitude-and phase-noise.Figure 5d-5f present in turn the respective diagrams for the wavelength-converted signal at the output.As observed, the wavelength-conversion process is successful despite the fact that it enhances the amplitude jitter and the phase variation of the signal and reduces the Q-factor of the decoded stream.Table 2 quantifies the above described observations using as a conventional metric for the amplitude-noise the ratio of the standard deviation of the power at the mark level (σ 1 ) over the mean power at the mark level (μ 1 ), and as a conventional metric for the phase-noise the standard deviation of the phase difference (σ phase ) between successive pulses around the ideal phase-difference values (0, 90°, 180°, 270°).Wavelength-conversion enhances the amplitude noise from 0.08 to 0.12, the standard deviation of the phase noise from 6.8° to 8.4°, and reduces the Q-factor of the decoded I-and Q-components from 16.9 down to 14.9 dB and from 17.1 down to 15.3 dB, respectively.Despite this degradation, the study indicates the feasibility of the technique, as this appears to yield wavelength-converted QPSK signals at acceptable quality levels.

Experimental setup
Figure 6 illustrates the experimental setup for the investigation of the MFPC performance with OOK and PSK signals.Two continuous wave (cw) signals at 1551 nm (λ 1 ) and 1558 nm (λ 2 ) were multiplexed and inserted into an electro-absorption modulator (EAM) that was driven by a 22 GHz sinusoidal signal for pulse carving with 6 ps pulse-width.The output of the EAM was amplified and fed into an arrayed-waveguide grating (AWG) to demultiplex the two wavelengths.The optical clock at λ 1 was forwarded into an InP monolithically integrated IQ modulator that was capable of intensity or phase modulation, depending on the bias settings.The IQ modulator was driven by the 2 31 -1 long pseudo-random bit sequence (PRBS) without differential encoding, yielding a 22 Gb/s OOK signal or a 22 Gb/s PSK signal.Either a phase-(PM) or an intensity-modulator (AM) in cascade imposed the respective type of noise on the PSK signal.The distorting Fig. 6.Experimental setup for OOK and PSK regeneration using the MFPC.modulators were driven by a 5.2 GHz sinusoidal signal with various peak-to-peak values resulting in different degrees of distortion.To study the performance of the MFPC with ASEnoise loaded signals, the distorting modulator was replaced by a variable attenuator that could adjust the OSNR of the input signal.
The second optical clock at λ 2 was driven into a nonlinear compression stage to acquire a pulse-width of 4.6 ps, and it was subsequently forwarded to serve as the input signal for the on-board SOA-MZIs.Polarization controllers were used at the input of the MFPC for both data and clock signals for adjusting the power splitting ratio of the DIs' couplers in order to maximize the ER of the decoded signals, as well as for biasing the SOA-MZI gate in order to achieve optimum switching performance.For operation with OOK signals, the data and the clock signals entered the MFPC through ports 2 and 9 with 0.73 mW and 0.38 mW average power, respectively.The output signal was taken in turn from port 20, it was filtered and further driven for bit-error rate (BER) measurements after a 4:1 electrical demultiplexing stage.For operation with PSK signals, the data and the clock signals entered the MFPC through ports 3 and 9 with 8.6 mW and 0.29 mW average power, respectively.The wavelength-converted PSK signal was taken from port 20, and was launched back to the device through port 27 in order to be decoded by the lower 22 GHz DI (see Fig. 2a).The complementary decoded streams were taken out from ports 13 and 14 and were forwarded to the evaluation unit.In all cases, the current in the SOAs was constant at 300 mA and the operating temperature was maintained at 22.5 degrees Celsius by means of an integrated thermo-electric cooler (TEC) with 320 mA current consumption.

Experimental results and discussion
The evaluation of the MFPC performance was based on BER measurements and observation of the eye-diagrams at the input and the output of the device.

Received power (dBm)
Received power (dBm) -log (BER)   Finally, Fig. 8 presents the corresponding eye-diagrams for these measurements referring to the undegraded inputs as the back-to-back (B-2-B) case, the degraded inputs, the regenerated outputs and the respective decoded streams in the case of PSK signals.For OOK signals with ASE-noise, the quality enhancement is associated with suppression of the peakpower fluctuations at the mark-and the space-level.For PSK signals with phase-noise only, the regenerator suppresses sufficiently the specific type of noise and allows for a decoded signal with higher extinction ratio, but slightly enhanced peak power variation due to the partial conversion of the phase-into amplitude-noise [6,11,14,15].For PSK signals with amplitude-noise only, the regenerator suppresses the peak power variations in the PSK and the decoded streams enabling the overall enhancement of the quality.Finally, the case of OSNR degradation resembles more the case of amplitude distortion, and thus the regeneration is achieved again mainly due to the suppression of peak power fluctuations.These conclusions are in good agreement with the simulation results and similar conclusions from previous studies on phase-incoherent regeneration techniques [6,11,14,15].
The operation speed of the MFPC for phase encoded signals is determined by the freespectral range (FSR) of the on-chip DIs, whereas for simple OOK signals it is limited only by the speed capabilities of the SOA-MZI gate.It is expected that operation up to 44 Gbaud is supported for all modulation formats owing to the short gain recovery of the SOA devices.Further increase of the operational speed can be supported by introducing design modifications on-chip such as the integration of higher FSR DIs and additional delay elements and variable optical attenuators allowing for push-pull operation [16].An additional enhancement for the MFPC design would be the integration of DIs with different FSR values hence enabling apart from multi-format signal processing, also selectable symbol rate operation.

Conclusions
We have presented a thorough investigation of the wavelength conversion functionality with regenerative capabilities of a hybridly integrated photonic circuit able to process signals with variable modulation formats and bitrates.Extensive simulation results reveal the potential of the MFPC to achieve all-optical regeneration of OOK and DPSK signals at 22 Gb/s and efficient wavelength conversion of DQPSK signals at 44 Gbaud, for a degraded input with different types of added noise (phase-noise, amplitude-noise and ASE-noise).Experimental demonstration at 22 Gb/s confirmed the regenerative properties for OOK and DPSK signals achieving power penalty improvement up to 1.5 dB.

Fig. 2 .
Fig. 2. (a) Layout of the multi-format processor, (b) blueprint of its silica motherboard, and (c) packaged and pigtailed device used in the experimental studies.

Figure
Figure 2a presents the layout of the MFPC illustrating only 28 of its 32 ports for clarity.The processor contains 16 input and 16 output pigtailed optical ports, 4 DIs, a monolithic array of 4 Indium-Phosphide (InP) SOAs that are placed within 2 MZIs, 15 phase shifters and a complex network of waveguides and over 30 couplers for operation and monitoring purposes.The MFPC can accommodate simultaneous processing of 2 OOK signals at variable data rate, 2 DPSK signals at 22 or 44 Gb/s or processing of a single DQPSK signal at 44 Gbaud, with the symbol rate of the phase modulated signals determined by the free-spectral range (FSR) of the on-chip DIs.The network of the waveguides allows for the signals to enter

Fig. 3 .
Fig. 3. Q-factor improvement at the output of the regenerator for: (a) OOK signals with different types and degrees of input signal degradation, and (b) decoded DPSK signals for different types and degrees of input DPSK signal degradation.

Fig. 4 .
Fig. 4. Upper row: eye-diagrams of the 22 Gb/s OOK signals at (a) the input, and (b) the output of the regenerator, corresponding to the maximum Q-factor improvement in Fig. 3a.Lower row: eye-diagrams of the decoded DPSK signals at (c) the input, and (d) the output of the regenerator, corresponding to the maximum Q-factor improvement in Fig. 3b.

Fig. 5 .
Fig. 5. Upper row: (a) Eye-diagram of the degraded input DQPSK signal, (b) constellationdiagram of the degraded input DQPSK signal, and (c) eye-diagram of the corresponding decoded I-component.Lower row: respective diagrams for the wavelength-converted QPSK signal at the output.

Fig. 7 .
Fig. 7. Measurements for operation with degraded OOK and PSK signals: BER curves for (a) OOK signals with ASE-noise, (b) PSK signals with ASE-noise, (c) PSK signals with phasenoise, and (d) PSK signals with amplitude-noise.

Figure 7
Figure7summarizes the results for OOK signals with ASE-noise and PSK signals with phase-, amplitude-and ASE-noise.Each BER curve in Fig.7acorresponds to the worst performing between the four 5.5 Gb/s electrically demultiplexed tributaries, whereas each curve in Fig.7b-dcorresponds to the worst performing between the two complementary decoded streams and their four tributaries, which are associated with the 22 Gb/s PSK signal.The performance variation both between the tributaries and the decoded streams in the case of the PSK signal was negligible, though, due to the absence of patterning effects and the symmetry of each SOA-MZI gate, respectively.No patterning effects were observed at the SOA-MZI output owing to the short SOA gain recovery time and the use of the short-pulse optical clock signal as a common input to the SOA-MZI erasing memory effects induced by the carrier density change of the SOA dynamics.For ASE-noise loaded OOK signals with 18.3 dB OSNR, power penalty improvement of approximately 1.5 dB at the level of 10 9 BER was achieved.For degraded PSK signals, the corresponding power penalty improvement was approximately 1.5 dB for signals with phase noise, approximately 1.1 dB for signals with amplitude noise and approximately 1.2 dB for signals with ASE-noise corresponding to an input OSNR value of 18.4 dB.

Fig. 8 .
Fig. 8. Experimental eye-diagrams of the back-to-back (left column) signals, the degraded input signals (middle column) and the regenerated output signals (right column): (a) OOK signals with added ASE noise, (b) PSK signals and corresponding decoded signals for ASE noise, (c) PSK signals and corresponding decoded signals for phase noise, and (d) PSK signals and corresponding decoded signals for amplitude noise.Time scale: 10ps/div.