We propose a novel scheme of OTDM utilizing pulse position modulation, where optical null headers (ONH) are inserted between the signal pulses periodically to allow channel identification. The ONH also achieves in-band clock distribution through the generation of high contrast pilot tone on the signal power spectra, enabling baud-rate flexible clock recovery. Using the novel scheme, clock recovery with a timing jitter of less than 200 fs is achieved at different baud rates up to 344 Gbaud. We demonstrate stable clock recovery with channel identification in 344-Gb/s OTDM transmissions over dispersion managed 3-km SMF.
© 2013 OSA
Optical time division multiplexing (OTDM) is an attractive technology enabling ultra-high speed (e.g. Tbit/s) data transmissions on a single carrier far beyond the speed limit of electrical devices . One drawback of using OTDM may be its limited reach as compared with multi-carrier transmissions at the same capacity. Therefore, OTDM is suitable for short-reach yet high-capacity applications such as local area networks (LANs) of future broadcasting stations, where uncompressed ultra-high definition (UHD) video signals have to be transmitted frequently without any technical constraints . The real time transmission of uncompressed UHD video signals in fact requires a bandwidth of 72 Gb/s or even higher. While such high-speed transmission can be realized by OTDM technologies, the switching of such high-speed signals is going to be the bottleneck. Because the conventional electrical IP based switching will no longer scale due to its too large power consumption at higher speeds, an alternative technology is sought in order to drastically reduce the power consumption of the networks that handle such huge-capacity video contents . We then proposed that switching in the optical domain be the solution, and called the network based on this concept “dynamic optical path network (DOPN)”.
In order to employ OTDM in DOPN based LANs, stable clock recovery and channel identification are indispensable. To date, many methods of recovering the clock from OTDM signals at a specific baud rate have been demonstrated. They are categorized into two groups. The first group recovers RF clocks by phase-locked loops (PLL) utilizing ultra-fast phase comparators based on optical nonlinearities in semiconductor devices [4–6], periodically-poled lithium niobate  or nonlinear optical fibers [8, 9]. The other group recovers optical clocks through injection-locking of mode-lock lasers  or optoelectronic oscillators . While these methods recover the clock at the original base rate from the received OTDM signals, they do not consistently specify the phase of the recovered clock with respect to the timing of each of the OTDM tributaries. This is because conventional OTDM signals carry no channel information. On the other hand, dynamic channel identification is essential in DOPN, because receivers need to de-multiplex dynamically received OTDM signals immediately after path switching correctly assigning signal channels. Therefore, conventional OTDM without the information of signal channels cannot be applied to DOPN, and the dynamic channel identification of OTDM has been an unsolved issue.
To allow dynamic channel identification in OTDM, it is necessary that channel information is superposed on the data by some means. Transmitting channel information is essentially equivalent to distributing in-band clock because the in-band clock can have a definite phase with respect to the OTDM tributaries, almost unaffected from the transmission through a fiber. To the best of our knowledge, only a few works have been reported on the channel identification in OTDM [12, 13]. Incorporating the mechanism of channel identification into OTDM generally complicates the system configuration . So far, we have proposed a simple scheme to achieve channel-identifiable clock recovery in on-off-keying (OOK) based OTDM by exploiting signal phase . Recently, we presented an expanded scheme based upon pulse position modulation (PPM), which is independent of modulation formats and can achieve fast yet robust clock recovery using conventional electronics. Using the new scheme, “dynamic optical path switching” was successfully demonstrated in the 172-Gb/s OTDM transmission of uncompressed UHD video signals . We call this “dynamically switchable OTDM (DS-OTDM).” It is worth noting that DS-OTDM does not require any additional components for its implementation, and that in principle, DS-OTDM is scalable in operating bit rate. In this paper, we present experimental study on the performance of the clock recovery with channel identification achieved by the DS-OTDM. The baud-rate flexible operations are confirmed at 43 Gbaud, 172 Gbaud and 344 Gbaud. The stable clock recovery is successfully demonstrated in 344-Gbaud OOK-based DS-OTDM transmission over 3-km SMF. Such baud-rate flexible operations will receive much importance in the context of “flexible grid based elastic optical network”, in which the bandwidths flexibly change in accordance with the traffic demands without sacrificing spectral efficiency .
2. Principle of DS-OTDM
The simplest way of implementing PPM in OTDM is to insert a small temporal gap between the OTDM pulses periodically. Figure 1 shows an example where N channels of parallel data are converted to a serial data channel by DS-OTDM. In this case, small temporal gaps are inserted between the signal pulses of ch.1 and ch. N. The timing of the small gap, which we call “optical null header (ONH)”, serves as a reference to identify signal channels. Before we implement DS-OTDM operations, we need to calibrate the gate timings of the corresponding OTDM channel at the receiver: Each channel signal is de-multiplexed at different gate timings. It is easy to find the position of ONH because the de-multiplexed signal diminishes when the gate opens at the timing of ONH, then channels from #1 to #N can be found consecutively as the gate timing is increased. By recording the timings of de-multiplexing each signal channel and ONH in advance, it becomes feasible to identify the channel number.
The insertion of ONH generates a pilot tone on the power spectra of the OTDM signal. The transmitted pilot tone enables the receiver to recover the clock signal using a standard PLL without relying on ultra-fast gate devices based on optical nonlinearity. Therefore, fast yet robust clock recovery can be achieved. Since the pilot tone is generated by ONH, the phase of the recovered clock is always fixed to the timing of each OTDM tributaries. This feature enables automatic and reproducible de-multiplexing of desired channels when optical path is switched in DOPN . The proposed scheme is applicable to phase-shift-keying (PSK) format as well as OOK format and is easily scalable to high baud rate signals by increasing the number of OTDM tributaries. In addition, this scheme is easy to implement because ONH can be inserted through adjustment of the pulse separation without using any additional components.
3. Basic performance of the clock recovery in DS-OTDM
3-1. Baud rate flexible operation
First, we verified baud-rate flexible operations by performing clock recovery from 172-Gbaud and 344-Gbaud DS-OTDM signals. The two DS-OTDM signals were generated from a 43-Gb/s RZ-OOK signal in PRBS (231-1) by the split-and-delay method as shown in Fig. 2. In both signals, ONH with a width of 1.5 ps was inserted between the pulses with a base clock rate of 43 GHz. This ONH corresponded to an overhead occupying 5.6% of the transmitted signal. Hereafter, we call the width of ONH relative to the cycle period of the base clock ( = 23.2 ps) as ONH ratio. A pair of variable optical attenuator and an EDFA was used for adjusting the signal power and OSNR. The observation using an electrical spectrum analyzer (ESA) showed that ONH successfully generated a pilot tone on the power spectra of the DS-OTDM signals. To observe the power spectra, optical signals with a power of 0 dBm were received by a photo-receiver, whose output was amplified by a narrowband RF amplifier with a gain of 24 dB. Figure 3(a) shows the power spectra of the two OTDM signals around the clock frequency of 43.018 GHz. The inset shows the temporal waveform of the signal observed with an optical sampling oscilloscope (OSO). For comparison, power spectra of the 43-Gb/s RZ signal are also shown in Fig. 3(a). In Fig. 3(a), the floor of the measured spectra, which was −98 dBm, was caused by the electrical noise of ESA. The powers of the 43-GHz pilot tone generated in 172-Gbaud and 344-Gbaud DS-OTDM signals were −24.8 dBm and −29.2 dBm, respectively. These powers were more than 20 dB smaller than the pilot tone generated in the 43-Gb/s RZ signal. They could be larger by increasing the width of ONH. However, it was not necessary because the pilot tone of the DS-OTDM signal had high contrast exceeding 70 dB. Actually, as shown in Fig. 3(a), all the three signals had similar power spectra around the clock frequency. This feature made it possible to realize stable clock recovery at any baud-rate. In practice, the pulse height of the OTDM signal was temporally unstable and could be uneven as shown in the inset. The uneven pulse height also generated a pilot tone at 43 GHz, whose power could be as large as −48 dBm. The pilot tone generated by 5.6% ONH was 18 dB larger than this spontaneous clock component and allowed clock recovery with a low timing jitter sufficient for performing stable de-multiplex operations. Besides these measurements, it was observed that pilot tone generated by ONH was approximately 4 dB larger in the case of PSK format than in the case of OOK format. This means that experimental verification obtained from the OOK signals also supports the validity of DS-OTDM in the PSK format.
To recover clock signal with low phase noise from the transmitted pilot tone, we chose to use PLL rather than spectral filtering. The PLL that was used in the experiment consisted of a dielectric resonant oscillator (DRO), whose frequency was locked to the 43-GHz pilot tone with a servo bandwidth of approximately 600 kHz. The locking time was less than 10 μs . Using this PLL, the clock was recovered from the three baud-rate signals of 43, 172, and 344 Gbaud. Figure 3(b) shows the RF spectra of the recovered clock. It is seen that the spectra of the recovered clock well reproduce the profile of the power spectra of the received signal. The spectra outside the servo bandwidth reflect that of the free running DRO. The three spectra shown in Fig. 3(b) have slightly different profiles. This is because overall gain of the PLL was not exactly the same in the three cases when the traces were recorded. The timing jitter of the recovered clock was roughly estimated from the measured spectra by integrating the phase noise from 2 kH to 2 MHz with the results of 52 fs, 47 fs and 51 fs for the signals of 43 Gbaud, 172 GBaud and 344 Gbaud, respectively. The fair agreement between the values obtained from different baud-rate signals proves the baud-rate flexibility of the DS-OTDM.
3-2. Performance at different operating conditions
While ONH plays an important role in DS-OTDM, it essentially reduces the transmission efficiency or, equivalently, spectral efficiency slightly. In principle, the width of ONH can be smaller than the width of one bit slot as long as the generated pilot tone has sufficiently high contrast. In order to determine the minimum ONH width, timing jitter of the clock recovered from 344-Gbaud OTDM signal was measured for various widths of ONH. The rms timing jitter of the recovered clock was measured by analyzing the time domain waveform using a wideband sampling oscilloscope (SO: Agilent 86100C DCA-J) with an electrical bandwidth of 80 GHz. The SO was equipped with a precision time base module (Agilent 86107A) and triggered by a 10.75GHz reference signal that was synchronized to the original clock. In this measurement, temporal resolution was limited to 200 fs by the jitter noise floor of SO. In Fig. 4, the measured rms timing jitter is plotted as a function of ONH ratio together with the power of the pilot tone generated by ONH. The inset shows the temporal waveform of the recovered clock observed with SO. The ONH occupied 12.5% overhead when its width was equal to one bit slot, which was 2.9 ps. To perform stable de-multiplexing operation, the timing jitter of the recovered clock is desired to be approximately less than 1/14 of the bit slot width, which is 208 fs at 344 Gbaud . The measurement results showed that clock recovery with timing jitter of less than 200 fs was achieved when ONH ratio was larger than 5%. As shown in Fig. 4, the power of the pilot tone decreases with ONH ratio. When ONH ratio was below 5%, the influence of spontaneous clock component could not be ignored and the timing jitter increased with the reduction of ONH ratio. For ONH ratio of less than 3%, stable locking was not achieved.
The minimum width required to ONH depends on the timing jitter tolerance required to the recovered clock. For the timing jitter to be less than 208 fs, the pilot tone generated by ONH needed to be more than 12 dB larger than the spontaneous clock component. Therefore, for stable de-multiplexing of 344-Gbaud signals, the minimum ONH width is estimated to be one third of the bit-slot (ONH ratio > 4.5%). As the generation efficiency of the pilot tone is nearly baud-rate independent, the present results suggest that clock recovery at 640 Gbaud or even higher is feasible. In this case, wider ONH width would be necessary because the requirement for the timing jitter becomes severer with the baud rate. We also expect that DS-OTDM is useful in optical Nyquist TDM transmissions , where the conventional scheme of clock recovery may be problematic. If one of the eight signal channels is switched off or greatly attenuated in 344-Gbaud optical Nyquist TDM transmission, 12.5% ONH is generated. This should ensure stable clock recovery.
The aforementioned measurement showed that stable clock recovery can be realized by ONH that is much narrower than one bit slot. This is for sure as long as the contrast of the pilot tone is high enough. Actually, when the optical signal to noise ratio (OSNR) of the signal was reduced, the contrast of the pilot tone decreased, leading to increased phase noise of the recovered clock. The effect of OSNR on the performance of clock recovery was studied for 344-Gbaud DS-OTDM signal with 6% ONH. In Fig. 5(a), the rms timing jitter of the recovered clock is plotted as a function of OSNR defined by 0.1 nm noise bandwidth together with the power of the generated pilot tone. The inset shows the temporal waveform of the recovered clock. Stable clock recovery with the timing jitter of less than 200 fs was achieved when OSNR was larger than 20 dB. For OSNR below 20 dB, it is seen that timing jitter increased with the decrease of OSNR. When OSNR was decreased to 7 dB, the timing jitter was increased to 420 fs. At this level, the signal waveform did not exhibit clear eye openings as shown in Fig. 5(b) and the signal quality became the main issue rather than the clock recovery. In optical fiber communication, various impairments cause signal distortion limiting the transmission reach. The present results suggest that stable clock recovery can be achieved even from the deteriorated DS-OTDM signal if the initial waveform is preserved.
4. 344-Gbaud DS-OTDM transmission over 3-km SMF
In order to further confirm the validity of DS-OTDM, we demonstrated 344-Gbaud OOK-OTDM transmission over a dispersion managed 3-km SMF span. The relatively short transmission distance was chosen assuming the future application of DS-OTDM in the LANs. The experimental setup is shown in Fig. 6. A mode-locked fiber laser (MLFL) operating at 1551 nm generated a 43 GHz pulse train with a pulse width (FWHM) of 0.85 ps. To reduce the pulses width, we performed wavelength conversion by degenerate four-wave mixing (FWM). For this purpose, the output from MLFL was mixed with a CW laser at 1565 nm and launched into a 100-m highly nonlinear fiber (HNLF), denoted as HNLF1 in Fig. 6. HNLF1 had zero-dispersion wavelength (ZDW) at 1530 nm and nonlinear coefficient of 12 W−1km−1. The conversion efficiency of FWM was −22 dB. The converted pulses at 1537 nm, which had a pulse width (FWHM) of 0.75 ps, were data-coded by a LN intensity modulator in a PRBS of 231-1. The base 43-Gb/s RZ-OOK signal was multiplexed to a 344-Gbaud OTDM signal by split-and-delay method. By adjusting the delay of each tributary pulse, a small gap was created between the DS-OTDM pulses in every eight pulses while the eight pulses in series were arranged with equal time intervals. The 344-Gbaud DS-OTDM signal was launched into a dispersion-managed transmission line composed of 3-km SMF and 426-m DCF. The inset of Fig. 6 illustrates the time domain waveform of the transmitted signal. The ONH had a width of 1.2 ps, occupying 5.2% overhead in the 344-Gbaud DS-OTDM signal.
The transmitted signal was amplified and de-multiplexed by FWM in a 100-m HNLF, denoted as HNLF2. HNLF2, which had ZDW at 1537 nm and nonlinear coefficient of 14 W−1km−1, generated idler centered at 1573 nm with a conversion efficiency of −19 dB. A part of the received signal with a power of 0 dBm was split for clock recovery. Using the recovered clock at 43 GHz, a mode-locked laser diode (MLLD) at 1555 nm was driven to generate pump pulses for FWM. The output pulse from MLLD was 2.5 ps (FWHM) wide and was compressed to 0.8 ps using a comb-like-profile fiber . The relative timing between the signal and the pump pulse was adjusted by a variable delay line. After HNLF2, the converted signal at 1573 nm was selected by a band-pass filter. We performed bit error rate (BER) measurement for the de-multiplexed 43-Gb/s signals, where the transmitted signal had a fixed OSNR of 32 dB.
Figure 7(a) shows the temporal waveform of the received 344-Gbaud DS-OTDM signal. Figure 7(b) shows the BERs measured for the eight tributaries. The received power was fixed at −5.5 dBm. The channel number was identified according to the manner described in section 2. It is seen that almost the same performances were achieved for all the tributaries. Figure 7(c) shows the BER curves measured for one tributary (ch. 6) as a function of the power received by a photo detector. The measured BERs are well below the FEC limit and verify the successful clock recovery. As mentioned in the previous section, if the impairment of the transmission line is carefully mitigated, the scheme is also applicable to long reach transmissions.
When optical path is switched dynamically in DS-OTDM, desired channels can be de-multiplexed automatically immediately after the path switching . To ensure this automatic de-multiplexing operation for a long period, however, stabilization of gate timing is necessary. This can be achieved through active control of delay using an error signal obtained from the de-multiplexed signal as demonstrated in Ref. 9.
We presented a concept and experimental verifications of DS-OTDM, which employed ONH to enable baud-rate flexible clock recovery and channel identification. The baud-rate flexible operation was verified through the achievement of clock recovery using a conventional PLL at different baud rates of 43-, 172- and 344 Gbaud. The recovered clock had a timing jitter of less than 200 fs in all cases. Measurements at various operating conditions showed that 4.5% of ONH ratio was sufficient to achieve stable clock recovery at 344 Gbaud unless OSNR is crucially low. We demonstrated stable clock recovery and channel identification through the BER measurement in 344-Gbaud OOK DS-OTDM transmission over dispersion managed 3-km SMF.
Part of this work was supported by Special Coordination Funds for Promoting Science and Technology of MEXT.
References and links
1. 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]
2. K. Oyamada, T. Nakatogawa, and M. Nakamura, “Ultra-high-definition television and its optical transmission,” IEICE Trans. Commun. E94-B, 876–883 (2011).
3. S. Namiki, T. Kurosu, K. Tanizawa, J. Kurumida, T. Hasama, H. Ishikawa, T. Nakatogawa, M. Nakamura, and K. Oyamada, “Ultrahigh-definition video transmission and extremely green optical networks for future,” IEEE J. Sel. Top. Quantum Electron. 17(2), 446–457 (2011). [CrossRef]
4. D. T. K. Tong, K. L. Deng, B. Mikkelsen, G. Raybon, K. F. Dreyer, and J. E. Johnson, “160 Gbit/s clock recovery using electroabsorption modulator-based phase-locked loop,” Electron. Lett. 36(23), 1951–1952 (2000).
5. O. Kamatani and S. Kawanishi, “Prescaled timing extraction from 400 Gb/s optical signal using a phase lock loop based on four-wave-mixing in a laser diode amplifier,” IEEE Photon. Technol. Lett. 8(8), 1094–1096 (1996). [CrossRef]
6. R. Salem, A. A. Ahmadi, G. E. Tudury, G. M. Carter, and T. E. Murphy, “Two-photon absorption for optical clock recovery in OTDM networks,” J. Lightwave Technol. 24(9), 3353–3362 (2006). [CrossRef]
7. L. K. Oxenløwe, F. Gómez-Agis, C. Ware, S. Kurimura, H. C. Hansen Mulvad, M. Galili, H. Nakajima, J. Ichikawa, D. Erasme, A. T. Clausen, and P. Jeppesen, “640-Gbit/s Data transmission and clock recovery using an ultrafast periodically poled lithium niobate device,” J. Lightwave Technol. 27(3), 205–213 (2009). [CrossRef]
8. J. H. Lee, S. Ohara, T. Nagashima, T. Hasegawa, N. Sugimoto, K. Igarashi, K. Katoh, and K. Kikuchi, “Clock recovery and demultiplexing of high-speed OTDM signal through combined use of bismuth oxide nonlinear fiber and erbium-doped bismuth oxide fiber,” IEEE Photon. Technol. Lett. 17(12), 2658–2660 (2005).
9. B. P. Kuo, A. O. J. Wiberg, C.-S. Bres, N. Alic, and S. Radic, “Ultrafast Clock recovery and sampling by single parametric devices,” IEEE Photon. Technol. Lett. 23(3), 191–193 (2011). [CrossRef]
10. T. Ohno, K. Sato, T. Shimizu, T. Furuta, and H. Ito, “Recovery of 40 GHz optical clock from 160 Gbit/s data using regeneratively mode-locked semiconductor laser,” Electron. Lett. 39(5), 453–455 (2003). [CrossRef]
11. D. Kong, H. Wang, Y. Li, J. Zang, S. Zhou, X. Jia, J. Wu, and J. Lin, “Ultra-low timing-jitter 40GHz clock recovery using EAM-MZM double-loop and its application in a 640Gbit/s OTDM system,” Opt. Express 20(26), B615–B623 (2012). [CrossRef] [PubMed]
12. M. Galili, H. C. Hansen Mulvad, L. K. Oxenlowe, H. Hu, E. Palushani, A. T. Clausen, and P. Jeppesen, “650 Gbit/s OTDM transmission over 80 km SSMF incorporating clock recovery, channel identification and demultiplexing in a polarisation insensitive receiver,” in Proc. Opt. Fiber Commun. Conf. 2010, OWO3, (San Diego, USA, 2010).
13. M. Kagawa, H. Murai, H. Tsuji, K. Sasaki, and K. Fujii, “Control and stabilization of bit-wise phase correlation in 160 (4 x 40) Gbit/s OTDM signal and its impact on transmission,” Opt. Express 16(14), 10039–10052 (2008). [CrossRef] [PubMed]
14. T. Kurosu, K. Tanizawa, and S. Namiki, “In-band clock distribution using signal phase for channel identifiable clock recovery in optical time-division multiplexing,” Electron. Lett. 48(14), 863–864 (2012). [CrossRef]
15. T. Kurosu, K. Tanizawa, S. Namiki, R. Akimoto, H. Kuwatsuka, T. Hasama, H. Ishikawa, T. Nakatogawa, K. Oyamada, Y. Tanaka, S. Ide, H. Onaka, and T. Asami, “Dynamic optical path switching in 172-Gb/s OTDM transmissions of ultra-high definition video signals using fast channel-identifiable clock recovery and integratable devices,” J. Lightwave Technol. 31(4), 594–601 (2013). [CrossRef]
16. M. Jinno, H. Takara, B. Kozicki, Y. Tsukishima, Y. Sone, and S. Matsuoka, “Spectrum-efficient and scalable elastic optical path network: architecture, benefits, and enabling technologies,” IEEE Commun. Mag. 47(11), 66–73 (2009). [CrossRef]
17. M. Jinno, “Effects of crosstalk and timing jitter on all-optical time-division demultiplexing using a nonlinear fiber Sagnac interferometer switch,” IEEE J. Quantum Electron. 30(12), 2842–2853 (1994). [CrossRef]
19. T. Inoue, H. Tobioka, K. Igarashi, and S. Namiki, “Optical pulse compression based on stationary rescaled pulse propagation in a comblike profiled fiber,” J. Lightwave Technol. 24(7), 2510–2522 (2006). [CrossRef]