Open Access
Add to BibSonomyAbstract
We propose an OTDM to WDM converter which enables wavelength tunability, flexible OTDM tributary to WDM channel mapping and modulation format transparency. The converted signals are obtained by four-wave mixing (FWM) the input 160 Gb/s OTDM signal with a multi-wavelength sampling pulse train (SPT). The generation of the multi-wavelength SPT starts by multicasting an optical clock signal. The multicast pulses are then individually delayed and reshaped by a programmable optical processor (POP), resulting in flexible generation of the SPT. Error-free performance was achieved in different OTDM tributary to WDM channel mappings. In addition, intermediate rate conversion (2x80 Gb/s) was also achieved simply by reconfiguring the POP.
©2012 Optical Society of America
1. Introduction
The latest research efforts in high capacity optical networks have been fueled by 100 Gb/s and 1 Tb/s Ethernet standards. In order to support such bit rates, the required capacity increase of optical links relied in the increase of the spectral efficiency. This applies to both wavelength division multiplexing (WDM) [1] and optical time division multiplexing (OTDM) systems [2]. While more than 1 Tb/s/ch has been already achieved in both systems, it is still uncertain how such systems can work together [3]. Nevertheless, the interface between both systems should certainly include multiplexing scheme conversion between OTDM and WDM [2–11].
OTDM single-tributary to WDM single-channel conversion has been implemented using different techniques [2–6]. Such techniques are based on different nonlinear subsystems such as nonlinear optical loop mirror [2], Mach-Zehnder interferometer semiconductor optical amplifier (MZI-SOA) [4], highly nonlinear fiber (HNLF) [5] and photonic chip [6]. OTDM multiple-tributary to WDM multichannel conversion can be achieved by parallelizing single-channel conversion techniques. However, parallel architectures present problems in synchronization and scalability [7]. As a result, alternative multichannel conversion techniques have been presented [7–11]. Even though such techniques rely on different operation principles, they have limitations concerning the wavelength tunability of the input or converted signals, OTDM tributary to WDM channel mapping, or modulation format transparency. The first two limitations are generally related with the generation of the sampling signals, whereas the latter depends on the used nonlinear process.
In this paper, we propose and experimentally assess an OTDM to WDM converter which addresses these limitations. The generation of the sampling pulses (SPs) is flexible both on time and wavelength domains thanks to the tunable frequency response of a POP. The SPs are mixed with the input OTDM signal by means of degenerate FWM in a HNLF, enabling modulation format transparent OTDM to WDM conversion.
2. Operation principle
The operation principle of the OTDM to WDM converter is depicted in Fig. 1 . Part of the input data signal is used to feed a clock recovery subsystem, which outputs a clean optical clock signal at the tributary rate. The optical clock signal is then multicast into N different wavelengths. N is the number of tributaries of the input OTDM signal, which are four in the illustration of Fig. 1. The resulting SPs are processed by a POP. Such device enables designing an arbitrary transfer function over a large frequency range within a given frequency resolution. The phase of the transfer function is adjusted to control the time delay imposed to each SP. In addition, the SPs can be reshaped through the adjustment of the POP’s amplituderesponse. This is very useful in compensating signal distortions caused by the wavelength multicasting system. The optimization of the frequency response of a POP with the purpose of producing SPs was already reported in [12]. However, only single-channel operation was investigated. After amplification and noise filtering the input data signal and SPs are launched in a HNLF. The N converted WDM signals are the product of FWM between the input data signal and SPs. The output tunable optical filter (TOF) filters the converted WDM signals.
3. Experiment
The experimental setup of the 160 Gb/s OTDM to WDM converter is shown in Fig. 2 . As the converter was tested in back-to-back (BtB), a clock recovery system was not required. A mode locked fiber laser (MLFL) provided a 40 GHz optical clock signal centered at 1539 nm. The full width at half maximum of the clock pulses was of 1.5 ps. The optical clock signal was split in two copies. One was used to produce a 160 Gb/s intensity modulated (IM) OTDM signal, and the other was fed to a wavelength multicasting system.
The optical clock signal was multicast by a multi-wavelength conversion scheme employing a MZI-SOA [13]. The optical clock signal was converted to three CW signals spaced by 600 GHz. The fourth CW signal, centered at 1564.42 nm, was not considered for two reasons. First, since the maximum output power of the erbium doped fiber amplifier (EDFA) 2 was of only 50 mW, the addition of the fourth CW signal would significantly degrade the power budget related with the FWM process in the HNLF. Second, the idler resulting from FWM in the HNLF would have a wavelength of 1517.21 nm, which is out of the gain bandwidth of the used C-band EDFAs. The MZI-SOA had a low yet observable polarization sensitivity. The polarization of each CW signal was adjusted while the other lasers were turned off with the purpose of maximizing the gain provided by the MZI-SOA. Such procedure resulted in approximately co-polarized SPs. The shape and pulse width of the SPs were optimized through the adjustment of the tunable optical delay lines (TODLs), variable optical attenuators (VOAs) and also by the phase shifters integrated in the MZI-SOA.
As shown in Fig. 3 (a1), the POP (Waveshaper 4000S) was firstly programmed to produce three rectangularly shaped bandpass filters, each with a bandwidth of 400 GHz, centered at the three CW wavelengths. The phase responses of the programmed bandpass filters centered at λ1, λ2 and λ3 were defined to achieve a constant group delay response of 12.50, 6.25 and 0 ps, respectively. The purpose was to obtain a SPT with wavelengths of λ3, λ2 and λ1. The optical spectrum and waveforms of each SP and of the resulting SPT are shown in Figs. 3(b1)-(f1). Each one of the SPs shown in Figs. 3(c1), (d1) and (e1) were obtained by activating only one of the three bandpass filters. All SPs have a pulse width of 3.5 ps. Both the group delay response and the waveforms show that the pulses are properly delayed by steps of 6.25 ps. However, the amplitude response of the POP shows that the insertion losses are different for each SP. This is explained by the tradeoff between insertion loss and added time delay, which is inherent to the POP’s operation principle [14]. Furthermore, the power of the SPs at the output of the MZI-SOA is not identical for the three wavelengths. The conjunction of thesenon-uniformities results in a SPT with unequalized amplitudes, as shown in Fig. 3(f1). The employed multicasting scheme has very attractive features such as inherent gain and possibility of individually tuning the wavelength of the SPs. Furthermore, there are no patterning effects as the input optical signal is not modulated. On the other hand, the SOAs add amplified spontaneous emission (ASE) noise to the SPs, which results in noise floors observed in Fig. 3(b1). In addition, the nonlinear response of the SOAs induces FWM between the high power input optical clock signal and SPs [13]. The FWM idlers can be observed in Fig. 3(b1). The idlers have the highest power on the signal centered at λ1, as it is the closest one to the input optical clock signal. As shown in Fig. 3(f1), this results in a thicker pulse trace which can ultimately impair the OTDM to WDM conversion.
Fig. 3 (a) frequency responses of the POP measured by an optical network analyzer, (b) optical spectra of the SPT, waveforms of the (c)-(e) SPs and (f) SPTs obtained for the POP configured to produce (up) rectangular bandpass filtering, (middle) line-by-line filtering, and (bottom) improved CR.
The impairments caused by the MZI-SOA on the SPs were mitigated by reconfiguring the POP. The insertion losses of the amplitude responses related with each SP were individually adjusted, thereby providing amplitude equalization of the SPT. The ASE noise floor and FWM idlers added by the SOAs were filtered out by setting the amplitude response of the programmed bandpass filters to maximum attenuation between the spectral lines of the SPs. The results are shown in Figs. 3(a2)-(f2). The group delay response of the POP could not be measured due to the discontinuities introduced in the amplitude response. Nevertheless, Fig. 3(f2) shows that the POP’s time delay response is correct, and that the amplitude equalization has improved. The second SP at λ2 falsely appears noisy and unequalized as the used large bandwidth optical sampling scope was able to capture the crosstalk between the pulse itself and the zero level of the neighboring pulses. The quality of the SPs has significantly improved, particularly at λ1. Figure 3(b2) shows that the noise floor and a considerable number of FWM idlers were suppressed. The idlers close to the spectral lines of the SPs could not be suppressed due to the limited frequency resolution of the POP. The minimum achievable –3 dB bandwidth of the bandpass filters centered at each spectral line was of about 7.5 GHz.
The SPs shown in Figs. 3(c1)-(e1) have a contrast ratio (CR) between 9.5 and 11 dB. The CR of the pulses shown in Figs. 3(c2)-(e2) improved by 2 dB. It is important to have a high CR of the SPs, as it directly affects the extinction ratio (ER) of the converted WDM signals. The CR can be improved by again reconfiguring the amplitude response of the POP. The CR of an optical clock signal mainly depends on the power of the central spectral line. Therefore, the CR of the SPs was improved by adjusting the insertion loss of the bandpass filters located at λ1, λ2 and λ3. The results are shown in the last row of Fig. 3. The CRs of the SPs at λ1, λ2 and λ3 improved to 15.7, 16.8 and 18.6 dB, respectively. Such a high contrast ratio obtained at λ3 resulted in amplitude ripple just before the pulse’s rising edge. The shape of the SPs could be further improved by individually optimizing each spectral line.
The SPs and input OTDM signal were amplified, filtered and then coupled into a HNLF with a length of 150 m, a nonlinear coefficient of 10.5 W−1km−1, zero dispersion wavelength of 1550 nm and a dispersion slope of 5 fs/nm2/km. The input powers of the SPs and OTDM signal were of 13 dBm and 21 dBm, respectively. OTDM to WDM conversion was obtained by degenerate FWM, which produces two sets of idlers. The set located at higher wavelengths enables modulation format transparent conversion, whereas the set located at lower wavelengths does not as the phase of each idler is twice the phase of the corresponding OTDM tributary. However, the idlers at higher wavelengths could not be considered as they lie outside the gain bandwidth of the available C-band EDFAs. Although the employed FWM scheme is polarization sensitive, it is possible to have polarization insensitive operation using the scheme proposed in [5]. The eye diagrams and optical spectra of the FWM idlers are shown in Figs. 4 (a2)-(a4) and Fig. 4(e), respectively. A FWM conversion efficiency of –20 dB was obtained. Even though the idlers are undistorted, the adjacent OTDM tributaries were not completely suppressed. This was mainly caused by the significant power of the zero level right before and after the SPs. The idlers were also impaired by the reduced roll-off of the TOFs used after the HNLF, tight channel spacing and noise added by the high power EDFA 1. Nonetheless, high Q-factors of 16.5 dB were measured in all converted signals.
Fig. 4 (a)-(d) SPTs and corresponding 40 Gb/s converted WDM signals located at (2) λo,1 = 1532.7 nm, (3) λo,2 = 1528.2 nm and (4) λo,3 = 1523.1 nm. The Q-factors are in dB. (e) Spectrum at the HNLF output. (f) BERs of the input and converted WDM signals. Inset: eye diagram of the BtB 40 Gb/s optical signal.
The phase response of the POP was readjusted to obtain a SPT with wavelengths of λ1, λ2 and λ3. The amplitude of the SPs was again equalized due to the changes made in the phase response of the POP. The SPs and obtained FWM idlers are shown in Figs. 4(b1)-(b4). Properly delayed SPs and undistorted converted signals were obtained. SPTs with wavelengths of λ3, λ2 and λ1 or λ1, λ2 and λ3 could be obtained simply by using a purely dispersive medium at the output of the MZI-SOA with a negative or positive dispersion sign, respectively. However, a purely dispersive medium cannot produce a SPT with an arbitrary arrangement of wavelengths, such as λ2, λ1 and λ3. The phase response of the POP was adjusted to achieve such SPT. The results shown in Figs. 4(c1)-(c4) demonstrate that the configured phase response did not degrade the SPs. As a result, undistorted converted signals were obtained. This demonstrates that the POP can produce a SPT with an arbitrary arrangement of wavelengths. The proposed OTDM to WDM converter can also be used to multicast one OTDM tributary to different wavelengths. This requires that all SPs are overlapped in time. The phase response of the POP was adjusted to add identical time delaysto all SPs, yielding overlapped SPs as shown in Fig. 4(d1). The measured Q-factors of the converted signals are significantly lower in comparison with the previous cases. The reason is that the crosstalk produced by the adjacent WDM channels now overlaps in time with the pulses of the converted WDM signal, resulting in a thicker pulse trace.
The bit error rates (BERs) of the converted optical signals at 40 Gb/s as function of the average power at the input of EDFA 3 are plotted in Fig. 4(f). Error-free operation (BER<10−9) was achieved in all signals. The power penalty of the converted signals relatively to BtB varies from 3.6 dB to 6.3 dB. The performance of the converted signals does not vary significantly among the different configurations. The highest power penalties are obtained for λo,1 since it is the converted signal closest to the high-power input OTDM signal. The power penalties obtained for λo,2 and λo,3 are identical. The converted signal at λo,2 has two adjacent converted signals, whereas the converted signal at λo,3 has only one. However, the central wavelength of the latter signal is located at the lower limit of the gain bandwidth of the used EDFAs. Even though low Q-factors were measured in Figs. 4(d1)-(d4), the corresponding BERs do not reveal a performance penalty. This can be explained by the bandwidth of 30 GHz of the photoreceiver used in the BER measurements. Such a low bandwidth resulted in the mitigation of the interchannel crosstalk previously captured by the optical sampling scope.
The repetition rate of the SPs can be doubled by suppressing half the spectral lines of each SP, leaving the remaining ones spaced by 80 GHz. The resulting SPs can be used to perform intermediate rate conversion, i.e., 160 Gb/s to 2x80 Gb/s WDM signals. The amplitude response of the POP was configured to achieve intermediate rate conversion, as shown in Fig. 5(a) . As only two WDM channels are required, the SPs centered at λ2 were suppressed. Figure 5(b) shows that the suppression of half the spectral lines of the SPs also provided additional suppression of the FWM idlers induced by the SOAs. Properly delayed, undistorted SPs were obtained, as shown in Figs. 5(c),(d). As a result, the converted optical signals presented an undistorted eye diagram with high Q-factors of 16.6 dB (λo,1) and 16.9 dB (λo,3).
Fig. 5 Results obtained for intermediate rate (80 Gb/s) conversion. (a) Frequency response of the POP, (b) optical spectra and (c)-(d) waveforms of the SPs and SPTs. (e) converted optical signals at λo,1 (up) and λo,3 (down).
4. Conclusion
We have proposed and experimentally assessed a 160 Gb/s OTDM to 4x40 Gb/s WDM converter based on FWM between the input optical signal and a multi-wavelength SPT. The SPs were obtained by multicasting a 40 GHz ultrashort pulse train into different wavelengths using a MZI-SOA. The multicast pulses were flexibly delayed and reshaped by a POP, which resulted in the suppression of ASE noise and FWM crosstalk, and also in the improvement of the CR. SPTs with different arrangement of wavelengths were produced, yielding undistorted converted WDM signals. All converted optical signals achieved error-free performance with a maximum power penalty of 6.3 dB relatively to BtB. Lower power penalties can be obtained by relaxing the channel spacing and by using TOFs with higher roll-off. The POP was also reconfigured to produce two 80 GHz time-interleaved SPs, enabling intermediate-rate conversion. Undistorted converted signals at 80 Gb/s with Q-factors higher than 16.5 dB were obtained.
Acknowledgments
This work was supported by the Fundação para a Ciência e Tecnologia project CONTACT (PTDC/EEA-TEL/114144/2009). The authors acknowledge Infractive and Finisar for loaning a Waveshaper 4000S.
References and links
1. M. S. Alfiad, M. Kuschnerov, S. L. Jansen, T. Wuth, D. van den Borne, and H. de Waardt, “11x224-Gb/s POLMUX-RZ-16QAM Transmission Over 670 km of SSMF With 50-GHz Channel Spacing,” IEEE Photon. Technol. Lett. 22(15), 1150–1152 (2010). [CrossRef]
2. H. C. H. Mulvad, M. Galili, L. K. Oxenløwe, H. Hu, A. T. Clausen, J. B. Jensen, C. Peucheret, and P. Jeppesen, “Demonstration of 5.1 Tbit/s data capacity on a single-wavelength channel,” Opt. Express 18(2), 1438–1443 (2010). [CrossRef] [PubMed]
3. L. K. Oxenløwe, M. Galili, H. C. Mulvad, H. Hu, H. Ji, J. Xu, E. Palushani, J. Areal, A. Clausen, and P. Jeppesen, “Ultra-high-speed optical signal processing of Tbaud data signals,” in Proc. ECOC’10(2010), paper Mo.1.A.1.
4. T. Hirooka, M. Okazaki, T. Hirano, P. Guan, M. Nakazawa, and S. Nakamura, “All-Optical Demultiplexing of 640-Gb/s OTDM-DPSK Signal Using a Semiconductor SMZ Switch,” IEEE Photon. Technol. Lett. 21(20), 1574–1576 (2009). [CrossRef]
5. H. Hu, H. C. H. Mulvad, M. Galili, E. Palushani, J. Xu, A. T. Clausen, L. K. Oxenløwe, and P. Jeppesen, “Polarization-Insensitive 640 Gb/s Demultiplexing Based on Four Wave Mixing in a Polarization-Maintaining Fibre Loop,” J. Lightwave Technol. 28(12), 1789–1795 (2010). [CrossRef]
6. T. D. Vo, H. Hu, M. Galili, E. Palushani, J. Xu, L. K. Oxenløewe, S. J. Madden, D.-Y. Choi, D. Bulla, M. D. Pelusi, J. Schroder, B. Luther-Davies, and B. J. Eggleton, “Photonic chip based 1.28 Tbaud transmitter optimization and receiver OTDM demultiplexing,” in Proc. OFC’10(2010), Post-deadline paper PDPC5.
7. C. Brés, A. Wiberg, B. Kuo, J. Chavez-Boggio, C. Marki, N. Alic, and S. Radic, “Optical Demultiplexing of 320 Gb/s to 8x40 Gb/s in Single Parametric Gate,” J. Lightwave Technol. 28(4), 434–442 (2010). [CrossRef]
8. A. H. Gnauck, R. M. Jopson, R. W. Tkach, C. J. McKinstrie, and S. Radic, “Serial-to-Parallel Demultiplexing Using WDM Sampling Pulses,” IEEE Photon. Technol. Lett. 21(2), 97–99 (2009). [CrossRef]
9. Y. Awaji, T. Miyazaki, and F. Kubota, “40 Gb/s WDM-Multicasting Wavelength Conversion From 160 Gb/s OTDM Signal,” Opt. Netw. Technol. 164, 292–298 (2005). [CrossRef]
10. G. Lei, C. Shu, and M. Fok, “All-Optical OTDM-to-WDM Signal Conversion Using Cross-Absorption Modulation With Time- and Wavelength-Interleaved Short Pulses,” IEEE Photon. Technol. Lett. 22(8), 571–573 (2010). [CrossRef]
11. K. J. Lee, S. Liu, F. Parmigiani, M. Ibsen, P. Petropoulos, K. Gallo, and D. J. Richardson, “OTDM to WDM format conversion based on quadratic cascading in a periodically poled lithium niobate waveguide,” Opt. Express 18(10), 10282–10288 (2010). [CrossRef] [PubMed]
12. M. A. F. Roelens, J. A. Bolger, D. Williams, S. J. Frisken, G. W. Baxter, A. M. Clarke, and B. J. Eggleton, “Flexible and Reconfigurable Time-Domain Demultiplexing of Optical Signals at 160 Gb/s,” IEEE Photon. Technol. Lett. 21(10), 618–620 (2009). [CrossRef]
13. N. Yan, J. del Val Puente, T. G. Silveira, A. Teixeira, A. P. S. Ferreira, E. Tangdiongga, P. Monteiro, and A. M. J. Koonen, “Simulation and Experimental Characterization of SOA-MZI-Based Multiwavelength Conversion,” J. Lightwave Technol. 27(2), 117–127 (2009). [CrossRef]
14. M. Roelens, S. Frisken, J. Bolger, D. Abakoumov, G. Baxter, S. Poole, and B. Eggleton, “Dispersion Trimming in a Reconfigurable Wavelength Selective Switch,” J. Lightwave Technol. 26(1), 73–78 (2008). [CrossRef]
