We simulate and experimentally demonstrate a novel all-optical clock recovery technique for a BPSK OFDM superchannel. Four-wave mixing in SOAs is used to strip the modulation from the superchannel sub-carriers, two of which are filtered and beat together in a photodiode to recover the clock.
© 2014 Optical Society of America
Modern optical communication networks require spectrally efficient schemes to meet ever-growing capacity demands. Transmission schemes using closely spaced frequency channels, called ‘Superchannels’, offer an attractive solution for capacities exceeding 400Gbit/s [1, 2]. The use of flexible grid wavelength-division multiplexing (WDM), with a granularity of <12.5GHz, has recently been suggested in the updated G.694.1 ITU-T recommendations (02/2012), demonstrating the demand for decreased channel spacing. All-optical orthogonal frequency division multiplexing (OFDM) is one format that could be easily used for flexi-grid implementations. It enables high capacity with high spectral efficiency , and with transmission impairments dominated by the baud rate of each optical subcarrier . Digital signal processing (DSP) is commonly used in the receiver for clock recovery, impairment compensation and reconstruction of the signal [1, 2], and DSP has also been suggested for use in an optical node . However, the availability of an analogue clock linked to the OFDM signal has many uses, including synchronization, phase tracking , and regeneration. Phase modulated formats such as Binary Phase Shift-Keying (BPSK) are commonly used in optical networks but recovering the clock from such carrier-less signals is challenging . However, by exploiting all-optical signal processing, such as four-wave mixing (FWM), in a nonlinear element, the carrier components may be recovered , which in turn may be used to recover the clock. Four-wave mixing is a well-known non-linear process and has been extensively used in many phase sensitive applications such as phase sensitive amplification  and phase sensitive frequency conversion [10–12].
We have recently reported a novel all-optical clock recovery scheme for a superchannel using FWM in semiconductor optical amplifiers (SOAs) . In this paper we report further experimental measurements and develop a numerical simulation of the technique to provide a detailed analysis. We recover the clock from a 53.54 Gbit/s BPSK OFDM superchannel and test the tolerance of this system both to degradations in the optical signal-to-noise ratio (OSNR) and to transmission over various lengths of single mode fibre (SMF). SOAs are the non-linear element of choice in this work, as they offer many advantages such as compactness and integration potential, whilst requiring only sub-mW input powers. Additionally they can provide net gain and are efficient nonlinear mixing devices over a range of wavelengths and difference frequencies .
2. Principle of operation
The all-optical clock recovery approach used in this experiment is shown in Fig. 1. Two adjacent subcarriers filtered from an optical OFDM superchannel with channel spacing of Δf and modulated with BPSK data at Δf are injected into an SOA along with a continuous wave (cw) pump. FWM between the cw pump and the filtered superchannel gives rise to two idlers located at 2fs1 – fP and 2fs2 – fP, where fs1 and fs2 are the carrier frequencies of each subcarrier and fP is the carrier frequency of the cw pump. These idlers will have double the frequency spacing (2Δf) and double the phase modulation (0 and 2π) of the original subcarriers . Since this is an identical phase relationship (ei0 = ei2π) these idlers will be stripped of any phase modulation. Beating these two idlers together in a photodiode will give a clock at their frequency spacing of 2Δf. This is double the original clock frequency used to produce the OFDM superchannel but can be easily down converted in the electrical domain to the original clock frequency, Δf. This clock recovery technique is similar in approach to the Viterbi-Viterbi method of recovering the carrier phase used in DSP .
A detailed description of the operation of FWM in SOAs can be found in . For the purposes of this paper, it is sufficient to view FWM in an SOA as a two-stage process, in which beating between the subcarriers and the cw pump modulates the carrier density in the SOA at frequency differences Δfm1 and Δfm2, shown in Fig. 1(a). This carrier density fluctuation then in turn modulates the amplitude and phase of all signals passing through the SOA. The two subcarriers, fs1 and fs2, are thus modulated by Δfm1 and Δfm2 respectively to produce sidebands at fs1 + Δfm1 ( = 2fs1 – fP) and fs2 + Δfm2 ( = 2fs2 – fP). This can be seen in the expanded idler view in Fig. 1(b). If the injected subcarrier at fs1 carries a phase modulation Δφ1 and the injected subcarrier at fs2 carries a phase modulation Δφ2, shown in Fig. 1(c), then Δfm1 will similarly contain the phase modulation Δφ1 and Δfm2 will contain Δφ2. As a result, the idlers produced are wavelength converted subcarriers each of which has been further phase modulated with the exact same phase information already encoded onto the subcarriers, i.e. fs1 + Δfm1 will carry the phase modulation 2Δφ1, as shown in Fig. 1(d).
Figure 1(b) shows that an idler is produced at fs1 + fs2 – fP, which one might think could also be used for clock recovery, but this is not the case in a real system. This idler is a combination of two beat components; the subcarrier at fs1 modulated by Δfm2 and the subcarrier at fs2 modulated by Δfm1; and the phase modulation of this idler is the sum of Δφ1 and Δφ2. Since the subcarrier data streams are decorrelated in a real system, this idler will not be stripped of its phase modulation, prohibiting its use in our clock recovery scheme.
The experimental setup is shown in Fig. 2. A distributed feedback laser (DFB), with a wavelength of ~1552.8 nm and a linewidth of ~300 kHz, was used in conjunction with two Mach-Zehnder modulators (MZMs) in series, each driven by a sinusoidal 10.7089 GHz clock signal from an oscillator, to produce a five channel comb. A delay interferometer, with a free spectral range of 21.3 GHz, was used as a disinterleaver separating the odd and even channels of the 10.7089 GHz spaced comb to separate output ports.
To properly mimic a real system several steps were taken to decorrelate all the subcarriers. Firstly, the odd and even channels were independently modulated using two MZMs. The odd channels were modulated with 10.7089 Gbit/s 231-1 PRBS BPSK data and the even channels were modulated with inverted 231-1 PRBS BPSK data at the same bit rate. To achieve further decorrelation between the odd and even channels, 10 m of SMF was added to the optical path of the odd channels. Both paths were then split using 50:50 couplers. Subsequently one path of each split was then delayed using 2m of SMF before all four optical paths were injected into a wavelength selective switch (WSS), as shown in Fig. 2. The WSS is a four port programmable filter which was used to select a different channel from each of the four input ports except port 4 where channel 1 and 5 were selected. Since each of the four optical paths have different lengths, all channels except the 1st and 5th were decorrelated with respect to each other . The WSS recombined these five channels together to produce a decorrelated 53.54 Gbit/s BPSK OFDM superchannel, as shown in Fig. 3(a) (black trace). All optical spectra in this paper are normalised to the original carrier frequency.
Unlike in , a polarisation beam splitter (PBS) was used to ensure that all subcarriers of the superchannel were co-polarised before transmission over a 50 km span of SMF followed by an erbium-doped fibre amplifier (EDFA). Another EDFA with no input was used to produce broadband amplified spontaneous emission (ASE), the power of which was controlled using a variable optical attenuator (VOA). This ASE was combined with the received superchannel in a 90:10 coupler before both were injected into a band-pass filter (BPF), as in Fig. 2. This method provided control over the OSNR of the received superchannel. In principle any two adjacent subcarriers may be filtered out using the BPF. Using two subcarriers at the edge of the superchannel will ease the filter bandwidth requirement, however, in this experiment we extracted the second and third subcarrier, shown in Fig. 3(a) (red trace). The pump ( + 3 dBm) was combined with the subcarriers (−8.2 dBm), (spectra are shown in the red trace of Fig. 3(b)) and passed through an isolator before injection into the SOA, which had a drive current of 400 mA. Since both the output power and FWM efficiency of the SOA are polarisation dependent, polarisation controllers (PC) were used to maximize the output idler components. At the output of the SOA, shown in Fig. 3(b) (blue trace), the two modulation-stripped idlers located at 2fS1 – fP and 2fS2 – fP were filtered out using a BPF and notch filter (green trace), before amplification and detection on a photodiode. The phase modulated idler due to the fS1 + fS2 – fP relationship is also visible, located between the two modulation-stripped idlers. The power envelope of the beating between the idlers in the photodiode occurs at their frequency separation and produces a sinusoidal clock signal at 21.4178 GHz. The output signal from the photodiode was passed through a narrowband electrical BPF, with a 1 dB pass band of ~4 GHz, to reduce noise. The output of the electrical BPF, captured on an RF spectrum analyser, is shown in Fig. 3(c).
The corresponding digital sampling oscilloscope (DSO) trace is shown in Fig. 4(a). The recovered clock (21.4178 GHz) had twice the original clock frequency (10.7089 GHz) but can be easily electrically down converted. The raw clock data measured on the RF spectrum analyser had a root mean square (RMS) phase noise of 0.072 radians (between 100 Hz – 10 MHz) and a signal-to-noise ratio (SNR) of 36.6 dB. The DSO image of the clock shown in Fig. 4(a) (blue data), indicates that a large amount of broadband noise is falling on the detector. A narrowband analogue bandpass filter, with a 3 dB bandwidth of 3 GHz, was applied in post-processing to the raw clock data to extract the clean 21.4178 GHz sinusoidal clock shown as the red solid curve of Fig. 4(a). A comparison of the original clock with the recovered filtered clock signal is shown in Fig. 4(b).
The input OSNR (0.1 nm bandwidth) of the entire BPSK superchannel was swept and the resulting SNR measured on the RF spectrum analyser was recorded. The OSNR here is defined as the ratio of the integrated power of the superchannel to the integrated power of the ASE for the superchannel’s optical bandwidth. A plot of the input OSNR vs. measured clock SNR is shown in Fig. 5(a). This clock recovery system was clearly tolerant to degradations in the received OSNR, which varied from 27 dB to 7.5 dB with less than 1 dB degradation in the clock SNR. Larger degradations were only observed for severely impaired OSNRs.
As two subcarriers are extracted using a BPF after the addition of ASE to the superchannel, any broadband ASE outside the bandwidth of this filter is suppressed. The total power into the SOA does not change with the addition of ASE; only the OSNR of the subcarriers is degraded. This means the FWM conditions in the SOA are unaffected, until a sufficiently small OSNR is injected. Since in a real system the required OSNR to detect a BPSK signal is >11 dBm , the OSNR, in this scenario, will never be low enough to significantly degrade the recovered clock SNR.
The performance of the clock recovery system as a function of FWM efficiency was also investigated by analysing the recovered clock SNR for various cw pump input powers, changing the pump-signal ratio. Figure 5(b) shows the overall results, where an optimum cw pump power between −3.5 dBm and −1 dBm was observed for an input signal power to the SOA of −9.8 dBm. Transmitting the superchannel over various lengths of SMF and measuring the corresponding clock SNR for constant cw pump and input signal powers to the SOA demonstrated the effect of dispersion. Recovered clock SNRs of 39.9 dB, 38.6 dB, and 36.6 dB were measured for back-to-back, 20km, and 50km of SMF respectively.
The all-optical superchannel clock recovery scheme was simulated using the numerical model described in . An OFDM superchannel was generated with five subcarriers spaced 10 GHz apart, each modulated with independent 2^10 De Bruijn BPSK patterns at 10 Gbit/s, to produce a 50 Gbit/s BPSK OFDM superchannel. A De Bruijn pattern was used instead of a PRBS pattern to help improve the decorrelation between the subcarriers. The De Bruijn pattern is identical to a PRBS pattern except that is one bit longer resulting in an even number of ones and zeroes. Decorrelation between subcarriers was achieved by inverting and delaying the De Bruijn pattern of each subcarrier with respect to the previous one. A dispersion of 1ns/nm was applied to the superchannel, representing transmission through ~59 km of SMF for a fibre dispersion of 17 ps/nm.km. Two of the subcarriers were filtered out using an analogue BPF and a plot of the superchannel and filtered subcarriers can be seen in Fig. 6(a). Due to the use of a De Bruijn pattern the optical spectrum differs slightly from that of PRBS BPSK signal, producing a slight dip where the carrier is suppressed.
A cw pump, located at −80 GHz from the centre of the superchannel, and the two filtered subcarriers were simulated in the numerical SOA model. The simulated output spectrum of the SOA can be seen in Fig. 6(b), where the produced idlers are clearly stripped of phase modulation, indicated by the spikes in the spectra. The two idlers produced from the 2fs – fP relationship, located at + 60 GHz and + 80 GHz, are extracted using an analogue BPF (highlighted in green in Fig. 6(b)) before they are combined and thereby beat to recover a clock at 20 GHz. An idler is also produced due to the fs1 + fs2 – fP relationship and is located at + 70 GHz. As mentioned previously this idler is dependent on the relative data streams of both subcarriers. The RF spectrum of the beating between these filtered idlers, shown in Fig. 6(d), shows a strong clock signal with ~20 dB SNR.
It is also possible to recover the clock without filtering just two subcarriers from the superchannel. Injecting the entire superchannel into the SOA along with a cw pump located at −140 GHz will modulation strip every subcarrier, as with the two channel case. The simulated output spectrum of the SOA can be seen in Fig. 6(c). Two of the modulation-stripped idlers are filtered out (highlighted in green) and the 20 GHz clock is recovered from them. There are two drawbacks to this approach. Firstly, a larger pump spacing (140 GHz) is required to ensure full separation between the superchannel and idlers at the output of the SOA. This will result in lower power idlers (~-4.5 dB), due to the dependence of the FWM efficiency in SOAs on the pump-signal frequency spacing . This is clearly seen when comparing Figs. 6(b) and 6(c). Secondly, the presence of extra subcarriers will produce more phase modulated idlers, similar to the fs1 + fs2 – fP component mentioned in section 2, which will overlap with the modulation-stripped idlers and will be seen as additional noise on the recovered clock. For example, the second idler, located at + 120 GHz in Fig. 6(c), will have noise added from the beating between the cw pump and the central subcarrier (located at 0 GHz) modulating the first subcarrier (located at −20GHz). However, when only two subcarriers are used the FWM produces only one phase modulated idler which is equidistant between the two modulation-stripped idlers. This minimises the additional noise at the photodetector.
In this paper we have experimentally demonstrated and simulated a novel all-optical clock recovery scheme for phase modulated superchannels. This system is capable of recovering a doubled clock frequency from a BPSK modulated optical OFDM superchannel using FWM in SOAs. A 53.54 Gbit/s BPSK superchannel, with 10.7089 GHz spacing between subcarriers, was transmitted over 50 km of SMF before extracting a clock signal at 21.4178 GHz, with a SNR of 36.6 dB and a RMS phase noise of 0.072 radians. This system was shown to be very tolerant to degradation of the input superchannel OSNR, as sweeping the OSNR from 27 dB to 7.5 dB reduced the recovered clock SNR by less than 1 dB. Transmission over various lengths of SMF minimally degraded the quality of the recovered clock, with only a 3.3 dB difference measured in the clock SNR between back-to-back and 50 km of SMF.
This work was supported by Science Foundation Ireland under 10/CE/I1853 (CTVR-II). The authors would also like to thank Dr. Frank Smith, from Pilot Photonics, and Dr. Jian Zhao for useful discussions.
References and links
1. Y.-K. Huang, E. Ip, Z. Wang, M.-F. Huang, Y. Shao, and T. Wang, “Transmission of spectral efficient super-channels using all-optical OFDM and digital coherent receiver technologies,” J. Lightwave Technol. 29(24), 3838–3844 (2011). [CrossRef]
2. X. Liu, S. Chandrasekhar, X. Chen, P. J. Winzer, Y. Pan, T. F. Taunay, B. Zhu, M. Fishteyn, M. F. Yan, J. M. Fini, E. M. Monberg, and F. V. Dimarcello, “1.12-Tb/s 32-QAM-OFDM superchannel with 8.6-b/s/Hz intrachannel spectral efficiency and space-division multiplexed transmission with 60-b/s/Hz aggregate spectral efficiency,” Opt. Express 19(26), B958–B964 (2011). [CrossRef] [PubMed]
3. D. Hillerkuss, R. Schmogrow, T. Schellinger, M. Jordan, M. Winter, G. Huber, T. Vallaitis, R. Bonk, P. Kleinow, F. Frey, M. Roeger, S. Koenig, A. Ludwig, A. Marculescu, J. Li, M. Hoh, M. Dreschmann, J. Meyer, S. Ben Ezra, N. Narkiss, B. Nebendahl, F. Parmigiani, P. Petropoulos, B. Resan, A. Oehler, K. Weingarten, T. Ellermeyer, J. Lutz, M. Moeller, M. Huebner, J. Becker, C. Koos, W. Freude, and J. Leuthold, “26 Tbit s-1 line-rate super-channel transmission utilizing all-optical fast Fourier transform processing,” Nat. Photonics 5(6), 364–371 (2011). [CrossRef]
4. F. C. G. Gunning, T. Healy, and A. D. Ellis, “Dispersion tolerance of coherent WDM,” IEEE Photonics Technol. Lett. 18(12), 1338–1340 (2006). [CrossRef]
5. P. J. Winzer, “An opto-electronic interferometer and its use in subcarrier add/drop multiplexing,” J. Lightwave Technol. 31(11), 1775–1782 (2013). [CrossRef]
6. P. Frascella, S. Sygletos, and A. D. Ellis, “A novel phase stabilization scheme for DPSK CoWDM signals using high order four wave mixing,” in European Conference and Exhibition on Optical Communication (ECOC) (2011), paper We.8.A.2. [CrossRef]
7. G. Contestabile, M. Presi, N. Calabretta, and E. Ciaramella, “All-optical clock recovery for NRZ-DPSK signals,” IEEE Photonics Technol. Lett. 18(23), 2544–2546 (2006). [CrossRef]
8. R. Weerasuriya, S. Sygletos, S. K. Ibrahim, F. C. G. Gunning, R. J. Manning, R. Phelan, J. O’Carroll, B. Kelly, J. O’Gorman, and A. D. Ellis, “Comparison of frequency symmetric signal generation from a BPSK input using fiber and semiconductor-based nonlinear elements,” IEEE Photonics Technol. Lett. 23(10), 651–653 (2011). [CrossRef]
9. R. Slavík, F. Parmigiani, J. Kakande, C. Lundstrom, M. Sjodin, P. A. Andrekson, R. Weerasuriya, S. Sygletos, A. D. Ellis, L. Gruner-Nielsen, D. Jakobsen, S. Herstrom, R. Phelan, J. O’Gorman, A. Bogris, D. Syvridis, S. Dasgupta, P. Petropoulos, and D. J. Richardson, “All-optical phase and amplitude regenerator for next-generation telecommunications systems,” Nat. Photonics 4(10), 690–695 (2010). [CrossRef]
11. R. P. Webb, J. M. Dailey, R. J. Manning, and A. D. Ellis, “Phase discrimination and simultaneous frequency conversion of the orthogonal components of an optical signal by four-wave mixing in an SOA,” Opt. Express 19(21), 20015–20022 (2011). [CrossRef] [PubMed]
13. M. J. Power, W. Jia, R. P. Webb, R. J. Manning, and F. C. Garcia Gunning, “Clock recovery of phase modulated optical OFDM superchannel,” in Optical Fiber Communication Conference (OFC) (2014), paper W3F.1.
14. A. J. Viterbi and A. M. Viterbi, “Nonlinear estimation of PSK-modulated carrier phase with application to burst digital transmission,” IEEE Trans. Inf. Theory 29(4), 543–551 (1983). [CrossRef]
15. J. Schroeder, L. B. Du, M. M. Morshed, B. J. Eggleton, and A. J. Lowery, “Colorless flexible signal generator for elastic networks and rapid prototyping,” in Optical Fiber Communication Conference (OFC) (2013), paper JW2A.44. [CrossRef]
16. P. J. Winzer and R. J. Essiambre, “Advanced modulation formats for high-capacity optical transport networks,” J. Lightwave Technol. 24(12), 4711–4728 (2006). [CrossRef]