1. Introduction

A clock extraction is one of the essential techniques for time-domain demultiplexing or data regeneration in future optical-time-division multiplexed (OTDM) transmission systems. Among various schemes reported so far, the phase-locked loop (PLL) is the most established technique for clock extraction. At bit rates higher than 40 Gbit/s, PLL with an optical/optoelectronic phase comparator provides a viable alternative to its electrical counterpart [1–4].

The straightforward method of phase comparison in the optical domain is based on the optical gate switching technique using a semiconductor optical amplifier (SOA) [1] or an electoabsorption modulator (EAM) [2–4]. However, such technique is unsuitable for 160-Gbit/s system applications due to the broad switching window of several ps. Although it is possible to reduce the switching window less than 3 ps by cascading SO As or EAMs, such a complicated configuration cannot be applicable to real systems [3].

In this paper, we propose a simple optoelectronic phase comparator using phase modulation and spectral filtering, which is free from ultrafast optical gate switching. Using our phase comparator, we construct a prescaled optoelectronic PLL (OE-PLL) and demonstrate extraction of a low-noise 10-GHz clock from a 160-Gbit/s OTDM signal.

2. Principle of operation

We consider the case where an electrical clock at the frequency of f _rep/4 is extracted from an optical pulse train with the repetition rate of f _rep. Figure 1 shows the schematic of our technique comparing the phase between the optical pulse train and the electrical clock, and Fig. 2 explains the principle of operation. Here, for simplicity, we assume a sinusoidal intensity modulation as the pulse train. The pulse train is phase-modulated at the same frequency as the RF clock, and the phase modulation index is π, as shown in Fig. 2 (a). Figure 2 (b) represents the optical spectrum before phase modulation. It is noted that the pulse-to-pulse phase characteristic peculiarly depends on the time delay Δτ of the optical pulse with respect to the phase modulation. We assume Δτ = 0 when peaks of the optical pulse train and the phase modulation coincide. Figures 2 (c) and (d) respectively show the temporal and spectral waveforms of the phase-modulated pulse train when Δτ = 0. It is similar to the carrier-suppressed return-to-zero (CS-RZ) pulse train since phase variations at peaks are “π, 0, π, 0”. On the contrary, when Δτ = 1/2f _rep, temporal and spectral waveforms are given by Figs. 2 (e) and (f), respectively. Phases at peaks are “π/√2 , π/√2, - π/√2, - π/√2” in this case. Note that the spectral component at the offset frequency of ±f _rep/2 is enhanced when Δτ = 0 (Fig. 2 (d)) while it is suppressed when Δτ =1/2f _rep (Fig. 2 (f)). Therefore, filtering the phase-modulated pulse train at the offset frequency of ±f _rep/2, we can obtain the phase error signal simply by measuring the output power of the filter.

 

Fig. 1. Schematic configuration of our optoelectronic phase comparator. f _rep is the repetition rate of the optical pulse train. P _in and P _out are the input power and output power of the bandpass filter (BPF), respectively. Δf is the bandwidth of BPF. PM: phase modulation.

Download Full Size | PDF

 

Fig. 2. Temporal and spectral waveforms of the optical pulse train after phase modulation. Δτ is the time delay of the optical pulse train with respect to phase modulation.

Download Full Size | PDF

When we employ the optical gate switching technique for clock extraction, the maximum bit rate is limited by the switching window width of the phase comparator. On the other hand, in our scheme, the maximum bit rate is definitely quad the phase modulation frequency; hence, our scheme is applicable to a 160-Gbit/s system by using a commercially available LiNbO₃ (LN) phase modulator with a 40-GHz bandwidth.

3. Performance evaluation of our phase comparator

Before clock extraction experiments, we evaluate the phase comparison characteristics between a 40-GHz optical pulse train and a 10-GHz electrical RF. This result is scalable to the phase comparison between a 160-Gbit/s optical signal and a 40-GHz RF.

First, a 40-GHz optical train with a 10-mW average power was generated by sinusoidal intensity modulation of a CW light at 40 GHz, and then it was phase-modulated through a LN modulator with the insertion loss of 5 dB. The modulator was driven by a 10-GHz electrical RF, whose power was adjusted such that the modulation index was π. Next, the phase modulated pulse train was filtered out at the 20-GHz offset frequency by an optical bandpass filter (BPF). The bandwidth of the BPF was 10 GHz, and the insertion loss was 18 dB.

Figure 3 shows measured optical spectra after phase modulation when Δτ = 0 and 1/2f _rep. The spectral components at ±20-GHz offset frequencies are suppressed when Δτ = 1/2f _rep, while the most of the pulse energy converge at the offset frequencies when Δτ = 0. Figure 4 is the ratio of the output power to the input power of BPF, P _out/P _in, measured as a function of Δτ together with the calculation result. We find that the agreement between experiment and theory is fairly well, and P _out/P _in is proportional to cos(2πf _repΔτ) with the dynamic range of over 9 dB. This fact suggests that our technique can work as a low-noise phase comparator. Although P _out was reduced to several-μW due to the surplus loss of BPF, it was possible to obtain the phase error signal with the peak-to-peak voltage over 100 mV by the use of a photodetector with a transimpedance of 10 kΩ. Such a high sensitivity is comparable to that of the conventional phase comparator using an electrical mixer.

 

Fig. 3. Measured optical spectra of the optical pulse train after phase modulation when Δτ = 0 and 1/2f _rep.

Download Full Size | PDF

 

Fig. 4. Output-input power ratio P _out/P _in of BPF in our phase comparator. L _BPF is the loss of BPF.

Download Full Size | PDF

4. Experiment of 10-GHz clock extraction from 160-Gbit/s OTDM signal

We perform 10-GHz clock extraction from the 160-Gbit/s signal using OE-PLL composed of our phase comparator, which compares the phase between a 160-Gbit/s optical signal and a 40-GHz RF.

Figure 5 is the configuration of the OE-PLL. The 160-Gbit/s OTDM signal with a 0-dBm average power was generated by intensity modulation of a 10-GHz optical pulse train with a pseudo-random data pattern at 10 Gbit/s followed by optical time-division multiplexing [5]. It was launched on the phase comparator consisting of a LN phase modulator (LN-PM) and a BPF with the bandwidth of 0.4 nm. The modulator was driven by a 24-dBm 40-GHz RF supplied from a 10-GHz voltage-controlled oscillator (VCO) through two frequency doublers. The phase modulation index was calculated to be π. The phase error between the 160-Gbit/s OTDM signal and the 40-GHz RF, which was obtained by photo-detecting the output light from the phase comparator, controlled the frequency of VCO. The offset DC voltage V _offset was added to the phase error signal, and an electrical low-pass filter (LPF) was used to suppress the spurious components of the phase error signal.

Figure 6 shows phase-noise spectra of the extracted 10-GHz clock and the original 10-GHz base clock of the 160-Gbit/s OTDM signal. We find no additional phase noise in the extracted clock in the offset frequency range below 100 kHz. Although a peak component was generated in the extracted clock at several hundred kHz, the additional timing jitter is calculated to be less than 50 fs.

 

Fig. 5. Experimental setup of 10-GHz clock extraction from a 160-Gbit/s OTDM signal by OE-PLL consisting of our phase comparator. V _offset is the DC offset of the phase error signal. LN-PM: LiNbO3 phase modulator, BPF: bandpass filter, LPF: low-pass filter, VCO: voltage-controlled oscillator.

Download Full Size | PDF

 

Fig. 6. Single side-band phase-noise spectra of the extracted 10-GHz clock and the original base clock of 160-Gbit/s OTDM signal.

Download Full Size | PDF

The quality of the phase error signal generated from our scheme depends on the phase coherency of the OTDM signal. Although the coherency of the OTDM signal was somewhat degraded with the delay line-based OTDM process, we still had a high-quality phase error signal whose dynamic range was over 8 dB. Such a high dynamic range is sufficient for stable phase locking. Indeed, the stable phase-locked state was maintained over 10,000 sec.

5. Conclusion

We have proposed the simple, high-stable prescaled OE-PLL consisting of a high-sensitive optoelectronic phase comparator, which is based on phase modulation and spectral filtering. Using the proposed OE-PLL, we demonstrate low-noise base-clock extraction from a 160-Gbit/s OTDM signal.