Open Access
Add to BibSonomyAbstract
A novel re-timing, re-amplifying, and re-shaping (3R) regeneration system is proposed to process multiple WDM (wavelength-division-multiplexing) channels simultaneously. Its re-timing capability is investigated by both simulation and experiment with polarization-scrambling method at 10 Gb/s bit rate. Jitter tolerance up to 0.8 UIpp is demonstrated with BER improvement and floor breaking ability.
©2006 Optical Society of America
1. Introduction
In fiber-optic communication systems, the fidelity of an optical signal is impaired by attenuation, dispersion and nonlinearities in optical fiber. To achieve long haul optical fiber transmission, “3R” regeneration has become a critical technology. The cost and complexity of current WDM regenerators increases in proportion to the number of wavelengths because separate OEO (optical-electrical-optical) conversion is required to regenerate each channel individually. All optical techniques toward 3R regeneration, such as SOA-based interferometers [1], EAM-based cross absorption modulators [2], and injection locking laser diode devices [3] have been exploited recently. Those approaches, though emphasizing high data rate and significant SNR improvement, still come short in scalability due to single channel processing. A new technique to regenerate multiple WDM channels using a single device is needed. In this paper, we demonstrate an all-optical 3R regenerator that regenerates multiple WDM channels simultaneously. Our novel approach relies on a patented all-optical sampling technology, based on the Terahertz Optical Asymmetric Demultiplexer (TOAD) [4].
The schematic of the simultaneous WDM 3R regenerator is shown in Fig. 1. The signal waveforms at various stages inside of the regenerator (A, B, C, and D) are illustrated in Fig. 2.
The input to the regenerator are 4×10 Gb/s WDM non-return-to-zero (NRZ) signals that are impaired by noise and timing jitter [see Fig. 2(a)]. First, dispersion compensation fiber (DCF) is used to compensate transmission dispersion and synchronize different delays experienced by the four WDM channels. Then TOAD 1 performs bit-parallel gating during which all four WDM channels are gated simultaneously, thus converting NRZ WDM signal into a perfectly re-timed return-to-zero (RZ) amplified signal [see Fig. 2(b)]. Only one TOAD is required for re-timing, independent of the number of WDM channels. The next stage performs bit-interleaving of the re-timed RZ signals. This stage utilizes a 1x4 thin film filter (TTF) demux -delay lines - 4x1 power combiner structure, to interleave, or “stagger” the four WDM channels [see Fig. 2(c)]. Using TOAD’s picosecond switching window, multiple WDM channels can be interleaved. The stage output feeds control port of TOAD 2 after passing an EDFA. Input port of TOAD2 is fed with multi-wavelength pulses spectral sliced from a supercontinuum source with minimal noise and extremely low timing jitter (about 25 fs). TOAD 2 serves as an optical modulator, using the interleaved WDM data from control port to gate the clean multi-wavelength pulses from the supercontinuum source at input port [see Fig. 2(d)]. With a fast-gain recovery SOA to provide the rapid gating, all four WDM channels can be simultaneous re-shaped. A typical supercontinuum source can generate spectrum as wide as 100nm, more than hundred of DWDM channels can be supported.
2. Enabling technologies
2.1 The TOAD
TOAD [4] was first demonstrated and patented by Princeton. Since then TOAD has been used in various applications for all-optical data processing [5–8]. The Sagnac version of the TOAD, shown in Fig. 3(a), uses an intraloop 2x2 coupler to inject control pulses to the nonlinear element (a polarization insensitive SOA). When the data signal enters the TOAD, it splits into equal clockwise and counterclockwise components which counter-propagate around the loop and arrive at the SOA at slightly different times. The control pulse arrives at the SOA just before the counterclockwise component of the signal, which is to be gated, but just after its clockwise counter propagating component and induces nonlinearities in the SOA. This cause the two components to experience different phase shifts, as shown in Fig. 3(b), and consequently recombine and exit the loop at the output port. All other data signals, for which the counter propagating components do not straddle the control pulse arrival at the SOA, exit the loop at the input port. The TOAD switching window (e.g. transfer function) is shown in Fig. 3(c), and its response can be described by the following equation:
where Gcw, Gccw, ϕcw, and ϕccw are the gain and phase experience by the clockwise and counterclockwise propagating signals respectively. The offset Δx defines the width of the switching window from the equation Δτ= 2Δxsoa/cfiber. If Δτ is set small (say a few picosecond), one can obtain optical signal processing at ultra-fast speed [5].
Fig. 3. (a) Schematic of a TOAD, (b) calculated phase evolution for CCW and CW pulses, and (c) the corresponding TOAD transmission
2.2 Regeneration capabilities and bit-parallel-gating of the TOAD
TOAD-based signal processing has shown very promising results. Kailight Photonics reported all-optical 2R on single channel using NRZ-data format [6]. In Ref. [7], optical data regeneration was demonstrated by removing timing jitter from one single channel. Experimental results showed that jitter amplitude up to 17ps (or 0.17 UIpp, unit interval peak-to-peak for 10Gb/s signal) can be eliminated with the TOAD-based regenerator. Using the ultrafast gating properties of the TOAD, three parallel OC-48 WDM channels after traveling over 27.5 km in optical fiber in NRZ format were re-synchronized. Bit error rates less than 10-9 were measured for each channel [8].
2.3 Supercontinuum generation
We generated supercontinuum by sending amplified optical pulses from a mode-locked EDFL into a piece of dispersion decreasing fiber (DDF), where spectral broadening will take place due to self phase modulation (SPM):
As the relation shows, the broadening of the phase δω is proportional to the pulse amplitude U. The DDF will help the broadening process even more due to the soliton compression relation N 2 = (γP0(T0)2 / |β2|) = 1. To maintain its low soliton order, N, the width of the optical pulse T0 will narrow because of the gradually decreasing |β 2| in the DDF. As the amplitude of the optical pulse in the process of compression increases further, high-order dispersion and nonlinearity comes into play and assist further spectral broadening. A set of WDM filters such as TFFs are used to slice out the designated WDM channels from the wide spectrum. A typical supercontinuum source can provide as much as a 100 nm of spectral bandwidth, thus capable of supplying over 100 DWDM channels.
3. Numerical simulation
To assist the characterization of the system parameters and the optimization of the 3R regenerator testbed, we choose VPI’s VPItransmissionMakerTM as our primary simulation platform. Starting out with a slightly modified Sagnac loop, the TOAD was constructed using two-way couplers and a bidirectional SOA. A continuous wave (CW) laser modulated with pseudorandom bit sequence (PRBS) NRZ signal was fed into the TOAD data input. The RZ clock control pulses were generated using an optical pulse generator. At the output of the TOAD a trapezoidal filter is used to filter out the clock signal.
To simulate the setup’s ability to remove timing jitter, a jitter generating module is added by applying a sine function to the built in “Electrical Jitter Module”. By altering the parameter “Jitter Sensitivity,” different amounts of jitter amplitude can be introduced to the NRZ data input. A white Gaussian noise module may also be added to simulate amplitude noise at the input. “Channel Analyzer” modules are attached to the input and output of the TOAD, which will analyze the signals by plotting signal waveforms, optical spectrums, eye diagrams, and bit error rate (BER) graphs. The simulation results of the re-timing TOAD are shown in Fig. 4.
Fig. 4. (a) Input and (b) output eye diagrams for 0.2 UIpp timing jitter. (c) BER curve plots for inputs and outputs at different jitter levels
Figure 4(a) and 4(b) shows the single channel signal waveform at the input and output port of the TOAD, respectively by putting 0.2 UIpp timing jitter on the input signal. The jitter removal effect of the TOAD is fairly dramatic, as the timing jitter is fully eliminated from the output waveform. Note that the data format is converted from NRZ to RZ. Different BER curves were compiled for input signals with jitter amplitude of 20 and 50 ps, (corresponding to 0.2 and 0.5 UIpp) as shown in Fig. 4(e). The BER curves at the output show improvement after retiming, giving a negative power penalty around -2 dB and -3 dB at 10-9 BER.
4. Experiment performance of 3R testbed
The implementation of the testbed follows the schematic in Fig. 1. The “retiming” TOAD 1 uses a 10Gb/s SOA from Alcatel, and a tunable optical delay line to control the width of the gating window. A mode-locked EDFL with 2ps FWHM and 10 GHz repetition rate was used as “bit-parallel gating control pulse.” Figure 5 shows the signal eye diagrams at various stages (See Fig. 1 and Fig. 2) in the testbed. Due to limited hardware, the four NRZ input channels are generated by modulating a combined output from four DFB cw laser using a 10 Gb/s LiNbO3 modulator with PRBS bit pattern. A typical eye diagram from one of the channel is shown in Fig. 5(a). The center wavelengths are 1539.77, 1541.35, 1542.94, and 1544.53nm at 200 GHz ITU grid. After bit-parallel gating of all four WDM channels by TOAD 1 (the control pulse was synchronized with the input 10 Gb/s data stream), the signals were converted into a perfectly re-timed parallel RZ formats having 10 ps FWHM, as shown in Fig. 5(b). These four channels were then spectrally separated using 1x4 TFF wavelength demultiplexer, appropriately delayed, and then multiplexed with 4x1 power coupler, thus effectively creating 40 Gb/s data stream. The TFF wavelength demultiplexer consists of four 200GHz TTFs with center wavelengths matching those of the DFB lasers. Figure 5(c) shows the eye diagram of all four WDM channels after the bit-interleaving stage. The interleaved 40 Gb/s optical signal is then used as the control for TOAD 2. The input port of TOAD 2 is fed by four time-interleaved clean RZ signals spectrally spliced from the supercontinuum source. Note that the 200GHz 1x4 TTFs demux used for spectral slicing have wavelengths centered at 1550.12, 1551.72, 1553.33, and 1554.94 nm. As the result the original WDM channels will be wavelength converted to these wavelengths at the TOAD 2 output. To avoid bit pattern affects, the SOA recovery speed of TOAD 2 in the proposed design must be faster than 40 Gb/s. At the time of experiments such a fast SOA wasn’t available to us as product. Therefore we proposed and implemented a parallel approach for this stage based on TOADs with high-confinement SOAs. The four WDM channels were handled by an array of four TOADs instead of one single TOAD 2.
Fig. 5. Signal eye diagrams at different stages thru the 3R regenerator: (a) one-channel NRZ signal at point A, (b) re-timed one-channel RZ signal at point B (the other three channels are turned off), (c) four channels displayed together after bit-interleaving at point C.
We investigated the timing jitter suppression performance of the regenerator. Timing jitter was superimposed on the input WDM signal using polarization scrambling technique utilizing a polarization scrambler and a piece polarization maintaining (PM) fiber. By scrambling the polarization of the input, jitter can be induced onto the signal due to the different indices of the two axes in the PM fiber [1]. The amount of timing jitter can be found by the equations Δτ =(1\cslow - 1/cfast) L. The same amount of jitter was applied to all four WDM inputs using one available polarization scrambling setup.
Fig. 6. The eye diagram of the NRZ input signal (a) when no jitter is applied and (b) with 0.5 UIpp jitter. The re-timed RZ signal eye diagrams are also shown in (c) and (d). (e) The BER traces at the input and output of the regenerator after re-timing.
Figures 6(a) and 6(b) displays the eye diagrams of the NRZ input signal with and without jitter, where the induced jitter amplitude is 0.5 UIpp. After retiming, the polarization-induced jitter is removed from the format-converted RZ signal, as shown in Fig. 6(d). BER measurements were performed at both the input and output of the regenerator, as shown in Fig. 6(e). The power penalty at the output in this case is about 1.5 dB. This is proved to be less than the power penalty at the input, which is about 2.5 dB, therefore it bears larger jitter tolerance. Due to the fact that our SOA is slightly polarization dependent, a potential improvement can be expected if the jitter isn’t induced by polarization scrambling.
The output BER curve still has a positive power penalty comparing to the input at 10-9 BER at Fig. 6(e). This is due to the fact that the applied jitter amplitude hasn’t reached the limitation of the receiver. The photoreceiver, an HP Lightwave Converter, also has different responses to NRZ and RZ signals and produces different slopes on the BER plot. To analyze the regenerative ability of the test bed even further, 80 ps of jitter was produced at the input by lengthening the PM fiber. From the BER plots shown Fig. 7(a), one can see that the receiver begins to have difficulty dealing with this amount of jitter and the input BER hits a floor around 10-6. After retiming, the output BER curve not only achieves negative power penalty, it also breaks the original floor and reaches 10-9, showing true signal improvement. In Fig. 7(b), where the jitter amplitude is increased to 84 ps, and the floor of the input BER rises up to around 10-3. The output BER in this case also achieves negative power penalty and lowers the floor to around 10-6. The reason for a floor showing up in output BER curve is because of the rising and falling edges of the input NRZ signal introduced by the data modulator. As the result, these edges will move into the TOAD gating window once the jitter amplitude is large enough, closing the eye at the output. The curves in Fig. 7(c) follow the same trends when the jitter was increased to 88 ps. Our experimental results indicate that the proposed scheme is capable of removing timing jitter from each individual parallel WDM channel as long as the maximum relative walk-off between any two of the “jittery” WDM channels is less than 80 ps (ensuring a clear eye overlap for sampling) by tuning the TOAD sampling position.
The respective RF signal eye diagrams measured at the output of the photo-receiver are shown in Fig. 8. The eye opening effect in each case is immediately seen, while the output RZ data signals inherit minor distortion outside the eye area due to the design and response of the photo-receiver (optimized for NRZ signal detection). This effect is not apparent when we measured the trace using the 30GHz Tektronix optical sampling scope [Fig. 6(d)].
Fig. 7. BER measurement plots from input and output of the 3R regenerator when the jitter amplitude is (a) 80 ps, (b) 84 ps, and (c) 88 ps, respectively.
Fig. 8. Converted RF signal eye diagrams measured at the input and output of the 3R regenerator when the jitter amplitude is (a)(c) 80 ps, and (b)(d) 88 ps, respectively.
5. Conclusion
A novel all-optical 3R regeneration system which is capable of processing multiple WDM channels using TOAD switching and supercontinuum spectral slicing is proposed. The operation of the regenerator was carried out both by simulation and experiment. We demonstrated the system functionalities and measured the performance of the test bed. Strong jitter-reduction ability of the regenerator was observed. Not only can the regenerator tolerate up to 0.8 UIpp in jitter amplitude with a negative power penalty, the output signal BER also breaks the floor obtained at the input. Instead of a single TOAD 2 an array of TOADs is used to handle the reshaping of signals in a parallel fashion to avoid the slow recovery speed limitations imposed by the current commercial SOAs. This limitation can be avoided with higher than 40Gb/s switching speed SOAs available which will also improve the scalability of proposed regenerator dramatically thus allowing a possibility of processing much higher count of WDM channels simultaneously.
Acknowledgments
We gratefully acknowledge the support of Defense Advanced Research Projects Agency (DARPA) for this work.
References and links
1. R. Inohara, K. Nishimura, M. Tsurusawa, M. Usami, and S. Akiba, “40 Gbit/s transmission with all-optical 3R regeneration using two-stage SOA-based polarization discriminated switch with assist light injection,” IEICE Trans. Electron. E87-C, 1106–1114 (2004).
2. M. Hayashi, T. Otani, H. Tanaka, and M. Suzuki, “Analysis on jitter tolerance of optical 3R regenerator,” IEEE Photon. Technol. Lett. 15, 1609–1611 (2003). [CrossRef]
3. J. Nakagawa, M. E. Marhic, and L. G. Kazovsky, “All-optical 3R regeneration technique using injection-locking in gain-switched DFB-LD,” Electron. Lett. 37, 231–232 (2001). [CrossRef]
4. J. P. Sokoloff, P. R. Prucnal, I. Glesk, and M. Kane, “A terahertz optical asymmetric demultiplexer (TOAD),” IEEE Photon. Technol. Lett. 5, 787–789 (1993). [CrossRef]
5. I. Glesk, J. P. Sokoloff, and P. R. Prucnal, “Demonstration of all-optical demultiplexing of TDM Data at 250 Gb/s,” Electron. Lett. 30, 339–430 (1994). [CrossRef]
6. E. Granot, R. Zaibel, N. Narkiss, S. Ben-Ezra, H. Chayet, N. Shahar, S. Sternklar, S. Tsadka, and P. R. Prucnal, “Tunable all-optical signal regenerator with a semiconductor optical amplifier and a Sagnac loop: principles of operation,” J. Opt. Soc. Am. B 22, 2534–2541 (2005). [CrossRef]
7. B. C. Wang, L. Xu L, V. Baby, D. Y. Zhou, R. J. Runser, I. Glesk, and P. R. Prucnal, “Experimental study on the regeneration capability of the terahertz optical asymmetric demultiplexer,” Opt. Commun. 199, 83–88 (2001). [CrossRef]
8. D. Y. Zhou, B. C. Wang, and R. J. Runser, “Perfectly synchronized bit-parallel WDM data transmission over a single optical fiber,” IEEE Photon. Technol. Lett. 13, 382–384 (2001) [CrossRef]
