Elsevier

Optics Communications

Volumes 298–299, 1 July 2013, Pages 213-221
Optics Communications

Comparison of nonlinear fiber-based approaches for all-optical clock recovery at 40 Gb/s

https://doi.org/10.1016/j.optcom.2013.02.053Get rights and content

Abstract

We compare two different nonlinear fiber-based approaches for all-optical clock recovery (AOCR) at 40 Gb/s. Both involve mode-locking a semiconductor fiber laser (SFL); one is based on the use of a nonlinear optical loop mirror (NOLM) while the second exploits nonlinear polarization rotation (NPR). We assess the impact of input signal characteristics on the RMS timing jitter and power fluctuation of the recovered clock. We also investigate the possibility for wavelength tunable operation. The results show that both approaches are capable of recovering a good quality clock signal over a large range of operating parameters. While the NPR-based approach has slightly better performance in terms of reduced power fluctuation and wavelength tunability, it is more sensitive to polarization. We also discuss the approaches for polarization-insensitive operation.

Introduction

The combination of coherent detection with spectrally-efficient modulation formats has allowed research systems to approach the maximum transmission capacity of single-mode optical fiber [1]. However, these systems make use of complex transmitters and receivers which are based on power-hungry electronics and digital signal processing. On the other hand, not all signal processing functions need to be performed in the electronic domain. Particularly, the use of all-optical signal processing in which signals are processed directly in the optical domain without the need for O/E conversion, may be one approach to reduce power consumption [2]. One specific signal processing function of importance is all-optical clock recovery (AOCR).

Hybrid opto-electronic approaches [3], [4], [5] have been investigated and are capable of operation at data rates of at least 40 Gb/s. However, they may be difficult to implement at higher data rates, e.g., 80 Gb/s or beyond. One approach to overcome these limitations is to use AOCR which has the advantages of high speed, simple configuration, and potential for providing increased functionality such as frequency down-conversion (subharmonic clock recovery) [6], [7]. Numerous techniques for AOCR have been demonstrated, including spectral filtering [8], [9], [10], self-pulsating laser diodes [11], [12], [13], [14], two-photon absorption in avalanche photodiodes [15], temporal Talbot effect in single-mode fiber (SMF) [16], and mode-locked fiber ring lasers [17], [18], [19], [20], [21], [22], [23], [24], [25], [26], [27], [28], [29]. Compared to the other approaches, mode-locked fiber lasers present several other advantages such as wide wavelength tunability and wide flexibility in terms of operating bit rate. The fiber lasers are based on an erbium-doped fiber amplifier (EDFA), a semiconductor optical amplifier (SOA), parametric amplification, or a combination of amplifying mechanisms. Different active and passive mode-locking mechanisms have been investigated, such as cross-gain modulation (XGM) or cross-phase modulation (XPM) in an SOA or length of fiber [17], [18], [19], [20], cross-absorption modulation (XAM) in an electro-absorption modulator (EAM) [21], [22], optical gating in a nonlinear optical loop mirror (NOLM) [23], [24], gain and/or phase modulation in a fiber optical parametrical oscillator (FOPO) [25], [26], nonlinear polarization rotation (NPR) in a length of optical fiber [27], the use of a linear optical amplifier [28], and a semiconductor saturable absorber [29]. Passive mode-locking techniques typically suffer from low repetition rates due to the relatively long cavity length and fundamental mode-locking. Moreover, generally speaking, SOAs or EAMs may impose limits on operating speeds due to their recovery/response times.

NOLMs are based on the ultrafast Kerr effect (∼fs) in optical fiber and have been used for many all-optical signal processing applications such as OTDM demultiplexing [30], modulation format conversion [31], all-optical logic gates [32], ultra-wideband (UWB) signal generation [33], and as an optically controlled gate for active mode-locking of fiber lasers [34], [35], [36]. In [36], a tunable mode-locked SFL incorporating a NOLM and synchronized to external periodic pulse trains at 40 GHz, 80 GHz, and 160 GHz with low root-mean-square (RMS) timing jitter was presented. Moreover, AOCR using a NOLM as the mode-locking element was demonstrated [24], [37]; however, previous reports involve operation at low data rates and no comprehensive analysis of AOCR was performed. In particular, the input signals were assumed to be ideal whereas in practical AOCR applications, we may expect transmission impairments to degrade the input signals.

A second ultrafast nonlinear effect in optical fiber, NPR, has also been used to demonstrate many all-optical signal processing functions including optical logic gates [38], OTDM tributary exchange [39], and UWB signal generation [40]. Recently, we demonstrated the use of NPR for mode-locking a SFL and for AOCR [27], [41].

In this paper, we compare the performance of AOCR at 40 Gb/s based on mode-locked SFLs using a NOLM and NPR. We assess the impact of average power, optical signal to noise ratio (OSNR), extinction ratio (ER), and differential group delay (DGD) of the input signal on the RMS timing jitter and power fluctuation of the recovered clock. We also investigate the wavelength tunability. Our results help to quantify the requirements of the input signal characteristics in order to obtain a recovered clock with a good performance and can be used to identify suitable nonlinear-based approaches for AOCR.

The remainder of this paper is organized as follows. Section 2 introduces the background and experimental setup of the two AOCR approaches, as well as the principle of operation. Section 3 presents the results on assessing the impact of different input signal characteristics on the quality of the recovered clock and compares the two approaches. In Section 4, we discuss the results and finally, in Section 5, we conclude.

Section snippets

NOLM

A NOLM is based on XPM in a length of fiber (e.g., highly nonlinear fiber, HNLF) which is placed in a Sagnac interferometer as shown in Fig. 1(a). Polarization controllers (PC1 and PC3) are used to adjust the state of polarization (SOP) of the pump Ep and probe (input) Ein signals, respectively, while PC2 is used to adjust the SOP of the NOLM. Two wavelength-division multiplexing couplers (WDMCs) are used to combine and separate the wavelengths of Ep and Ein. The input Ein is split into two

Experimental results

Fig. 6(a)–(d) shows the typical recovered clock and corresponding optical spectra from the two AOCR approaches when injecting an RZ-OOK signal (without any degradation) with an average power of 3 dBm; the SOA currents are 120 mA and 140 mA for the NOLM-based and NPR-based approaches, respectively.

For NOLM-based AOCR, the pulse width is 9.1 ps, the output wavelength is 1551 nm, and the 3 dB bandwidth is around 0.51 nm. The corresponding time-bandwidth product is 0.58, indicating that the pulses are

Discussion

In this paper, we have compared the use of a SFL mode-locked by a NOLM or NPR for AOCR at 40 Gb/s. Both approaches can recover a clock with an RMS timing jitter performance that allows for error-free operation (i.e., less than 1.5 ps [46], [47]) over a large range of input signal parameters. However, these two approaches cannot be used with input signals that have been degraded by a DGD of more than 7.3 ps (NPR-based) or 10.4 ps (NOLM-based) and do not work well when the OSNR of the input signal is

Summary

In summary, both NOLM and NPR-based AOCR approaches are promising in terms of providing low RMS timing jitter and power fluctuation over a range of input signal characteristics. For example, a large range of average input power can be used from −21 dBm to −6 dBm and AOCR can be achieved with input signals having an ER from 8.7 dB to 19 dB, with an OSNR higher than 15 dB, and with a DGD up to 10.4 ps (NOLM) or 7.3 ps (NPR). All the parameters and their corresponding ranges of the two approaches we

Acknowledgments

This work was supported by the Natural Science and Engineering Research Council of Canada (NSERC) via the CREATE program in Next-Generation Optical Networks.

References (51)

  • K.G. Vlachos

    Optics Communications

    (2003)
  • E. Ip et al.

    Optics Express

    (2008)
  • L.K. Oxenløwe et al.

    International Journal of Optics

    (2012)
  • F.G. Agis et al.

    IEEE Photonics Technology Letters

    (2006)
  • T. Ohara et al.

    IEEE Journal of Selected Topics in Quantum Electronics

    (2007)
  • W. He et al.

    Optics Express

    (2011)
  • E. Tangdiongga et al.

    IEEE Photonics Technology Letters

    (2004)
  • M. Silva et al.

    Journal of Lightware Technology

    (2011)
  • M. Jinno et al.

    IEEE Photonics Technology Letters

    (1990)
  • C. Giampiero et al.

    IEEE Photonics Technology Letters

    (2004)
  • C.H. Kouloumentas et al.

    IEEE Photonics Technology Letters

    (2006)
  • I. Monfils et al.

    Journal of Lightware Technology

    (2009)
  • X. Tang et al.

    IEEE Photonics Technology Letters

    (2008)
  • I. Kim et al.

    Journal of Lightware Technology

    (2005)
  • W. Mao et al.

    Journal of Lightware Technology

    (2002)
  • R. Salem et al.

    IEEE Photonics Technology Letters

    (2004)
  • D. Pudo et al.

    Journal of Lightware Technology Letters

    (2007)
  • K. Vlachos et al.

    IEEE Photonics Technology Letters

    (2000)
  • J. He et al.

    Optics Express

    (2005)
  • A.D. Ellis et al.

    Electronics Letters

    (1993)
  • L.F.K. Lui et al.

    IEEE Photonics Technology Letters

    (2007)
  • L.R. Chen et al.

    Journal of Lightwave Technology

    (2008)
  • S. Bigo et al.

    Electronics Letters

    (1995)
  • J.H. Cordero, L.R. Chen, M. Rochette, Proceedings of OECC, Sydney, Australia,...
  • J. Li et al.

    IEEE Journal of Selected Topics in Quantum Electronics

    (2012)
  • Cited by (7)

    View all citing articles on Scopus
    View full text