As channels rates in optical networks are expected to exceed 100Gb/s in the near future, new optical techniques for clock recovery will have to be developed for optical regeneration. This paper describes an optical clock recovery method based on a mode-locked laser diode. Experimental results show that a 42GHz high quality optical clock can be retrieved from a 170Gb/s OTDM data signal. Chirp transfer between the incident signal and the recovered clock signal is investigated using the SHG-FROG method. Results demonstrate that this clock recovery technique is invariant to input dispersion varying between ±75ps/nm, making it ideal for use in 3R regenerators.
©2008 Optical Society of America
In order to accommodate higher data rates and allow high-speed Internet and new broadband services, such as on-demand HD-TV and interactive gaming, future high capacity optical networks will be required to operate at channel data rates in excess of 100Gb/s. As these signals propagate across the network, pulse distortion due to chromatic dispersion and fibre nonlinearities can result in severe signal degradation, affecting overall system performance. Therefore in order to increase transmission distance and data rate, it is necessary to periodically regenerate the propagating data signal. In WDM systems, signal regeneration is carried out on individual data channels using electronic processing . This requires the optical signal to be first converted into the electronic domain, and then after regeneration, to be converted back into optical format so that it can continue on its journey across the network. As this technique relies upon the use of photodetectors and electronic clock recovery circuitry, it is limited by the maximum speeds at which current integrated electronic circuits can operate at, which presently is of the order of 40Gb/s. Thus, networks of the future will require regeneration techniques to be developed that can operate without the need for electronic conversion and processing. These new optical regenerators will therefore amplify, shape and re-time the signal entirely in the optical domain. It is anticipated that one of the key components of an all-optical regenerator will be optical clock recovery (OCR).
The work presented here describes an OCR method based on self-pulsation of a mode-locked quantum dot Fabry-Perot (ML-QD-FP) laser diode. The ML-QD-FP has been developed in the framework of the French research program ROTOR . In particular, this paper will investigate the performance of the OCR to high-levels of chromatic dispersion being imparted on a high-speed OTDM signal, with pulse durations in the order to 2–3ps. Experimental results will show that for a dispersion range of 150ps/nm, the group delay of the regenerated clock signal varied by less than 1ps/nm. The group delay is retrieved directly using the SHG-FROG technique.
This paper is divided as follows. Section 2 will briefly describe one current electronic clock recovery technique and mention some of the new optical clock recovery techniques that have been reported. Section 3 will explain the fundamentals behind OCR using self-pulsating laser diodes as this is the technique that is employed during the experiments described in Section 4. Section 5 presents and discusses the experimental setups and results obtained for the chirp transfer analysis of the regenerated clock, while Section 6 summaries the major findings of this work.
2. Optical clock recovery
Current clock recovery in optical networks is accomplished using an electrical circuit known as a phase locked loop (PLL). A conventional electrical PLL is composed of three basic components: a feedback loop, phase detector, and a voltage controlled oscillator (VCO). It operates by comparing the phase of the input signal and a local generated reference signal (VCO) within the phase detector, and using the error signal to raise or lower the frequency of the VCO until it matches the frequency of the input signal . Once synchronised, the frequency and phase of the input signal can be determined, allowing a locally generated clock signal to be generated. The major disadvantage associated with using a PLL is the maximum speed at which the phase comparator can operate at . This makes it unsuitable for operation in network where the data rate exceeds 40Gb/s per channel.
A number of different techniques have been proposed to replace the electrical phase comparator with a much faster nonlinear optical process that operates directly on the optical signal, resulting in an optical PLL. These methods include the use of gain modulation in a semiconductor optical amplifier (SOA) , amplitude modulation in an electro-absorption modulator (EAM) , and the nonlinear optical-to-electrical conversion process of Two-Photon Absorption (TPA) in a silicon avalanche photodiode . Even though all of these processes have demonstrated the recovery of a high quality optical clock signal from an incoming data signal operating in excess of 100Gb/s, all required the use of a locally generated clock signal (VCO and/or optical pulse generator) and feedback loop configuration. For practical implementation, a more compact and stable solution is preferred .
3. Optical clock recovery via self-pulsating laser diodes
The solution presented here is based on a mode-locked laser (MLL) and operates without the need of a PLL and high-speed RF electronics, at frequencies above which current commercial PLL can successfully operate . This makes it an ideal candidate for optical clock recovery in future high-speed optical networks.
Optical clock recovery using a passively mode-locked laser is accomplished by injecting the input data signal into a DC-biased self-pulsating laser. This incoming signal then modulates the amplitude or the phase of the light within the laser cavity. If the base-period (T, as shown in Fig.1) of the injected signal is equal to, or a sub-multiple of, the laser round-trip time, mode-locking of the laser occurs . This will result in the generation of a train of optical pulses from the MLL at a repetition rate determined by the lasers cavity length, and represents the recovered clock signal from the input data signal. Such a technique allows the clock recovery process to be a simple, compact and stable , while producing a high quality clock exhibiting a low temporal jitter  and optical noise .
One of the first demonstrations of optical clock recovery using a self-pulsating laser diode occurred in 1988 when a dual-section DFB laser was used to extract an optical clock signal from a 196Mb/s input signal . Since then, numerous others have demonstrated optical clock recovery at increasingly higher data rates. Recent work has confirmed successfully operation generating clock signals at frequencies ranging from 40GHz to 160GHz [14, 15], with 180GHz performance demonstrated using a multi-section DFB laser diode. . The work presented here uses passive mode-locking of a single chip, single section MLL to generate an optical clock at 42.68GHz from a 170.72Gb/s OTDM input signal. The MLL employed is a Fabry-Perot semiconductor laser made of a buried ridge structure with a QD active layer on InP substrate . By employing a QD active layer, the device offers a higher spectral purity when compared to bulk devices which leads to a reduction in the amount of phase noise present on the recovered clock signal. This in turn results in an overall reduction in the amount of jitter generated in the clock recovery process . The QD-FP laser is 1080μm-long, has a central lasing wavelength around 1570nm (L-band). Varying the current and temperature of the laser allows the locking of the self-pulsating signal to the repetition frequency of 42.68GHz, in a range of 20MHz, obtaining the clock extraction from the optical data signal.
4. Optical clock recovery
It has already been shown that by using the mode-locked quantum-dot Fabry-Perot (ML-QD-FP) laser diode described above, it is possible to successfully extract a high quality optical clock from input data signals operating with data rates ranging from 40Gb/s up to 160Gb/s . This work will concentrate on evaluating the performance of the OCR technique for varying the levels of chromatic dispersion imparted on the input OTDM signal. This analysis was carried out by investigating the transfer of chirp between the injected OTDM signal, operating at 170.72Gb/s, and the recovered clock signal at 42.68GHz. The chirp was retrieved by using a second harmonic generation (SHG) frequency resolved optical gating (SHGFROG)  setup to analyse the pulses entering (OTDM signal) and exiting (clock signal) the ML-QD-FP laser diode.
FROG analysis allows for the complete characterisation of ultrashort optical pulses. It uses the same Michelson interferometer as used in an autocorrelator except that the signal is spectrally resolved at each value of the temporal delay. This requires that the photomultiplier tube be replaced with a spectrometer, and that the temporal delay be provided by an electronically controlled translation stage. From the generated spectrogram it is possible to extract the complex amplitude (intensity and phase) of the optical pulse, without any assumption about its shape or structure. The FROG measurement scheme is advantageous over other pulse measurement schemes such as bandwidth limited oscilloscopes and traditional autocorrelation methods as its retrieves intensity and corresponding phase information . This property allows for the study of the dispersion transfer of the clock recovery process.
4.1 Experimental setup for optical clock recovery
Figure 2 shows the Optical Time Division Multiplexing (OTDM) transmitter that was used in the optical clock recovery experiments. It consists of an actively mode-locked pulse source (u2t TMLL) generating a train of optical pulses with durations less then 2ps at 1550nm. The TMLL operated at 10.67GHz, which corresponds to STM-64 bit-rate plus a 7% forward error correction (FEC) overhead. This signal was then modulated with a 27-1 PRBS data signal generated using a programmable pattern generator (PPG) and wideband data modulator. The 10.67Gb/s data signal was multiplied up to 170.72Gb/s using an optical multiplexer, which consisted of a four independently switchable stages with variable delay lengths within each stage. The high-speed OTDM signal was then amplified using an EDFA in order to overcome the insertion loss of the multiplexer. This signal was then monitored using a 1ps resolution optical sampling oscilloscope (OSO from PicoSolve Inc.).
The 170.72Gb/s signal then entered the OCR circuitry as shown in Fig. 3, which consisted of an EDFA, optical circulator, ML-QD-FP laser diode and optical filter. The self-pulsating ML-QD-FP laser diode had a peak emission wavelength around 1570nm (L-band). By varying the bias current and temperature of the device, the FP laser’s frequency was tuned close to 42.68GHz, thereby allowing an optical clock to be extracted from the injected signal . The optical power of the signal injected into the ML-QD-FP was 5dBm, and this remained constant during the course of the experiment for different levels of imparted dispersion. The recovered clock signal was then optically filtered (BW=5nm, centred at 1563nm) to eliminate any remnants of the input data signal which operated at 1550nm and to allow only a small part of the ML-QD-FP wide spectrum to pass. The filtered optical clock signal was then analysed using a SHG-FROG for chirp transfer characterization and a BER circuitry which was used to measure the quality of the recovered clock signal.
In order to verify that the ML-QD-FP laser diode was locking to the incident OTDM data signal, the output RF spectrum of the laser was examined on an electrical spectrum analyzer. Figure 4(a) shows the detected electrical spectrum of the output of the ML-QD-FP when there is no injected input signal (free running operation) resulting in an unlocked signal operating around 42.69GHz. Figure 4(b) shows the output spectrum when the 170Gb/s OTDM signal is injected in the laser diode. The spectral width and shape of the signal are consistent with the laser locking to a sub-multiple of the 170.72Gb/s input signal. The ML-QD-FP was designed to operate at 42.68GHz. The operating laser parameters have been chosen in order to recover the clock signal and are not optimized to achieve the best mode-locked spectral line width .
Previous work has already shown that the OCR described here can be used to successfully extract a high-quality optical clock from 40 and 160Gb/s OTDM data signals due to the sub-harmonic locking process . The main focus of this paper is to investigate the performance of the described clock recovery technique for varying levels of chromatic dispersion imparted on the 170.72Gb/s OTDM signal.
5. Chirp transfer characterization of ML-QD-FP clock recovery technique
As mentioned, the chirp will be measured using the SHG-FROG technique as this method allows the retrieval of the intensity and phase profile of the optical pulse. Knowledge of the chirp of the recovered clock signal is important if this clock signal is used for 3R regeneration as any chirp present will influence the evolution of the shape of the propagating optical pulse through fibre. Large amounts of chirp can lead to pulse broadening and signal distortion, adversely affecting system performance. In addition, the presence of chirp increases the time-bandwidth product, reducing spectral efficiency of the signal. Finally, the accurate knowledge of the chirp can be used to optimize dispersion compensation and/or pulse compression schemes.
5.1 Experimental setup for chirp transfer characterisation
The experimental setup used to investigate the chirp transfer of the OCR process is shown in Fig. 5. It consists of the same OTDM transmitter detailed in Fig. 2 except that the signal is unmodulated. In order to get the most accurate FROG measurement of the pulse width and the frequency chirp of the recovered optical clock signal, the PRBS data signal was changed to an input signal comprising of a dispersed continuous optical pulse train. The high-speed pulse train enters a tunable dispersion compensating module (TDCM, from TeraXion) which allowed the dispersion of the data signal to be varied between ±75ps/nm. The effects of dispersion were monitored using a 1ps resolution Optical Sampling Oscilloscope (OSO, from PicoSolve Inc.). The dispersed signal than entered the OCR circuitry already described in this paper. The generated 42.68GHz recovered optical clock signal is then analysed using the OSO and FROG technique.
5.2 Chirp characterization experimental results
In order to visualize the affects that various levels of dispersion has on a propagating data signal, Fig. 6 shows the eye diagrams of the 170.72Gb/s OTDM PRBS data signal. The eye diagrams shown were recorded using the OSO after the tunable dispersion compensating module under two different dispersion regimes; (a) Zero (0ps/nm) dispersion; (b) −75ps/nm dispersion. The pulse separation in Fig. 6(a) is 5.8ps corresponding to the multiplexed data rate, while the pulse width is approximately 3.2ps. This pulse duration is slightly high for a 170Gbit/s OTDM data signals and results from pulse broadening experienced in the EDFA before the TDCM. Fig. 6(b) displays the eye diagram for the same 170Gbit/s signal when the dispersion was set to −75ps/nm. As expected the eye is completely closed due to adjacent pulses overlapping.
Figures 7(a) & (b) shows oscilloscope traces of the 42.68GHz optically recovered clock for the two different dispersion regimes, 0ps/nm and −75ps/nm respectively. In both cases, the pulse separation is 23.4ps which corresponds to a repetition rate of 42.68GHz. The measured pulse duration on the OSO was in the order of 1.8ps for both cases shown, while the temporal jitter was <600fs for both signals. The average power of the recovered clock signal after the 5nm optical filter (as shown in Fig. 3) was 2mW (3dBm), corresponding to a peak optical power of 30mW, while the time-bandwidth product was 0.9.
Figures 8(a) &(b) shows the retrieved pulse intensity and chirp profile respectively from the FROG measurement of the 42.68GHz recovered clock signals for the two different dispersion regimes shown in Fig. 6 and Fig. 7. This Fig. shows that there is negligible difference in the intensity and chirp profile of the clock signal when the incident OTDM has no dispersion (0ps/nm) and when the dispersion is set to −75ps/nm. The measured pulse width was 1.5ps which is slightly less than that recorded using the OSO. This could be accounted for by the limited temporal resolution (approx. 1ps) of the OSO. It is interesting to note that the FROG analysis displays two pulse pedestals, positioned approximately 2.4ps away, and 15dB down, from the main peak. The formation of the pedestals shown in may be due to the combination of the non-ideal optical filtering and nonlinear chirp of the output signal from the MLLD chip. These pedestals are not clearly visible on the oscilloscope traces in Fig. 7 due to its limited dynamic range of the OSO. Figure 8(b) shows that the chirp profile of the recovered clock signal has a slight negative chirp across the central portion of the pulse.
The above dispersion analysis was repeated for a number of different dispersion values. Figure 9(a) plots the variation of the measured pulse width of the recovered clock signal as a function of varying input dispersion imparted on the data signal. There is only a slight variation (100fs) between the measured pulse duration of the clock signal across the entire dispersion range. Figure 9(b) presents a plot of the group delay measured across the pulse as a function of dispersion imparted by the TDCM on the input signal. As displayed, the group delay varies less than 1ps/nm over the entire dispersion tuning range of 150ps/nm.
In order to show the performance of the clock recovery technique to various levels of dispersion imparted on the input signal, the 42.68GHz clock signal generated was re-modulated with data and a BER analysis was carried out. This results of this analysis is shown Fig. 10, with the imparted dispersion varying between −51ps/nm and +35ps/nm. As can be clearly seen, there is negligible power penalty (~1dB) incurred over the entire tuning range. This helps to verify the results presented in Fig. 9.
This paper has presented an optical clock recovery method based on a ML-QD-FP laser diode. It was shown that using such a method, a 42.68GHz high-quality optical clock signal can be retrieved from a 170.72Gb/s OTDM data signal. An investigation into the effects of varying input dispersion levels has on the optical clock recovered signal was then carried out. The effects of dispersion were monitored by using the SHG-FROG method, allowing intensity and chirp profile of the signals to be characterized. Results presented show that for a dispersion range of 150ps/nm, the group delay of the output signal varied by less than 1ps/nm. This demonstrates that the OCR technique presented is insensitive to distortions of the injected signal caused by chromatic dispersion. In addition, the temporal jitter on the recovered clock signal was <600fs over the entire dispersion tuning range. These results show that a ML-QD-FP laser diode is an ideal candidate for optical clock recovery in the next generation of 3R regenerators. In addition to signal regeneration, OCR will find applications in other signal processing tasks such as all-optical demultiplexing .
References and links
1. S. A. Hamilton, B. S. Robinson, T. E. Murphy, S. J. Savage, and E. P. Ippen, “100Gb/s Optical Time-Division Multiplexed Networks,” J. Lightwave Technol. 20, 2086–2100 (2002). [CrossRef]
2. J. Renaudier, B. Lavigne, M. Jourdran, P. Gallion, F. Lelarge, B. Dagens, A. Accard, O. Legouezigou, and G. H. Duan, “First demonstration of all-optical clock recovery at 40GHz with standard-compliant jitter characteristics based on quantum-dots self-pulsating semiconductor laser” in Proceedings of 31st European Conference on Optical Communications (Glasgow, Scotland, 2005), pp.31–32.
3. K. Y. Lau, “Short-Pulse and High-Frequency Signal Generation in Semiconductor Lasers,” J. Lightwave Technol. 7, 400–419 (1989). [CrossRef]
4. M. Saruwatari “All-Optical Signal Processing for Terabit/Second Optical Transmission,” IEEE J. Sel. Top. Quantum Electron. 6, 1363–1374 (2000). [CrossRef]
5. O. Kamantani and S. Kawanishi, “Ultrahight-Speed Clock Recovery with Phase Lock Loop Based in Four-Wave Mixing in a Traveling-Wave Laser Diode Amplifier,” J. Lightwave Technol. 14, 1757–1767 (1996). [CrossRef]
6. J. P. Turkiewicz, E. Tangdiongga, G. D. Khoe, and H.de Waardt, “Clock Recovery and Demultiplexing Performance of 160-Gb/s OTDM Field Experiment,” IEEE Photon. Technol. Lett. 16, 1555–1557 (2004). [CrossRef]
7. R. Salem, A. A. Ahamdi, G. E. Tudury, G. M. Carter, and T. E. Murphy, “Two-Photon Absorption for Optical Clock Recovery in OTDM Networks,” J. Lightwave Technol. 243353–3362 (2006). [CrossRef]
8. Z. Hu, H-F. Chou, K. Nishimura, M. Usami, J. E. Bowers, and D. J. Blumenthal, “Optical Clock Recovery Circuits Using Traveling-Wave Electroabsorption Modulator-Based Ring Oscillators for 3R Regeneration,” IEEE J. Sel. Top. Quantum Electron. 11, 329–337 (2005). [CrossRef]
9. P. Rees, P. McEvoy, A. Valle, J. O’Gorman, S. Lynch, P. L,ais, L. Pesquera, and J. Hegarty, “A Theoretical Analysis of Optical Clock Extraction Using a Self-Pulsating Laser Diode,” IEEE J. Quantum Electron. 35, 221–227 (1999). [CrossRef]
10. K. Smith and J. K. Lucek, “All-optical clock recovery using a mode-locked laser,” Electron. Lett. 28, 1814–1816 (1992). [CrossRef]
11. V. Roncin, A. O’Hare, S. Lobo, Jacquette E., L. Bramerie, P. Rochard, W.-T. Le, M. Gay, J.-C. Simon, A. Shen, J. Renaudier, F. Lelarge, and G.-H. Duan, “Multi-Data-Rate System Performance of a 40-GHz All-Optical Recovery Based on a Quantum-Dot Fabry-Perot Laser,” IEEE Photon. Technol. Lett. 19, 1409–1411 (2007). [CrossRef]
12. V. Roncin, S. Lobo, L. Bramerie, P. Rochard, A. Shen, F. Van Dijk, G.-H. Duan, and J. C. Simon, “Demonstration of Chromatic Dispersion and Optical Noise insensitivity of a Quantum-Dash based Fabry-Perot Laser in All-Optical Clock Recovery at 40Gbit/s,” in Proceedings of 33rd European Conference on Optical Communications (Berlin, Germany, 2007), pp.165–166.
13. M. Jinno and T. Matsumoto, “All-optical timing extraction using a 1.5μm self-pulsating multielectrode DFB LD,” Electron. Lett. 24, 1426–1427 (1988). [CrossRef]
14. S. Arahira and Y. Ogawa, ”Retiming and reshaping function of all-optical clock extraction at 160 Gb/s in monolithic mode-locked laser diode,” IEEE J. Quantum Electron. 41, 937–944 (2005). [CrossRef]
15. I. Kim, C. Kim, G. Li, P. LiKamWa, and J. Hong, “180-GHz clock recovery using a multisection gain-coupled distributed feedback laser,” IEEE Photon. Technol. Lett. 17, 1295–1297 (2005). [CrossRef]
16. T. Ohno, K. Sato, T. Shimizu, T. Furuta, and H. Ito, “Recovery of 40 GHz optical clock from 160Gbit/s data using regeneratively modelocked semiconductor laser,” Electron. Lett. 39, 453–454 (2003). [CrossRef]
17. R. Trebino, K. W. DeLong, D. N. Fittinghoff, J. N. Sweetser, M. A. Krumbugel, and B. A. Richman, “Measuring ultrashort laser pulses in the time-frequency domain using frequency-resolved optical gating,” Rev. Sci. Instrum - American Institute of Physics 68, 3277–3295 (1997).