We demonstrate a single-channel 1.92 Tbit/s, 64 QAM coherent optical pulse optical time-division multiplexing (OTDM) transmission by utilizing frequency-domain equalization (FDE). FDE makes it possible to compensate precisely for the waveform distortions caused by hardware imperfections thus greatly improving the error vector magnitude (EVM) of the demodulated 64 QAM signal compared with that obtained with a conventional FIR filter. As a result, a coherent 64 QAM OTDM transmission over 150 km with a bit error rate of below the forward error correction limit of 2x10−3 (requiring 7% overhead) was achieved for the first time.
© 2013 Optical Society of America
The latest progress on 100 Gbit/s optical transmission technologies highlights the fact that expanding the transmission capacity toward 1 Tbit/s and beyond is becoming an important issue in optical communication research . Coherent RZ pulse transmission with a combination of multi-level modulation format and optical time-division multiplexing (OTDM) is an attractive candidate to achieve such a large capacity, since this makes it possible to realize Tbit/s transmission by using low speed electronic devices and with a modest single-channel symbol rate compared to conventional OTDM [2, 3]. This technology has already been used to demonstrate a single-carrier 5.1 Tbit/s, 16-quadrature amplitude modulation (QAM) coherent transmission at 10 Gsymbol/s × 64 OTDM . Moreover, a 10.2 Tbit/s, 16 QAM coherent signal transmission over 29 km with a self homodyne method has also been demonstrated by further increasing the OTDM multiplicity to 128 . Here, the transmitted signal was homodyne detected with part of the original non-modulated pulse signal.
Previously, we reported an increase in the QAM multiplicity to 32 and demonstrated an 800 Gbit/s, 32 QAM transmission at 10 Gsymbol/s × 8 OTDM over 225 km by utilizing an optical phase-locked loop (OPLL) in combination with a return-to-zero-continuous wave (RZ-CW) conversion scheme [6, 7]. At our coherent receiver, we compensated for waveform distortions caused by hardware imperfections using a time-domain equalization (TDE) technique that employed a finite impulse response (FIR) filter. However, the frequency resolution of FIR filters is inherently limited to several tens of MHz due to tap number limitations designed to avoid computational complexity. On the other hand, the frequency-domain equalization (FDE) technique, which allows an increase in the frequency resolution while maintaining relatively low computational complexity, has been applied to multi-level QAM coherent transmission [8, 9]. We have recently applied this technique to our coherent pulse OTDM transmission, and demonstrated the effectiveness of FDE with a 1.6 Tbit/s, 32 QAM (10 Gsymbol/s × 16 OTDM)-150 km transmission experiment .
In this paper, based on the performance improvement of waveform distortion compensation using FDE compared to FIR filter in OTDM-32 QAM transmission , we increased the QAM multiplicity from 32 to 64, and successfully transmitted a single-channel, polarization-multiplexed (pol-mux) 1.92 Tbit/s, 64 QAM coherent optical pulse OTDM signal over 150 km for the first time. The use of FDE greatly improved the error vector magnitude (EVM) of the demodulated 64 QAM signal compared with that obtained with a conventional FIR filter.
2. Experimental setup for single-channel 1.92 Tbit/s, 64 QAM coherent optical pulse transmission
Figure 1 shows the experimental setup for a 1.92 Tbit/s, 64 QAM coherent optical pulse transmission. The coherent optical source at the transmitter was a CW, C2H2 frequency-stabilized fiber laser . This laser operated at 1538.8 nm with a linewidth of 4 kHz. The laser output was fed into an optical comb generator , which consisted of a dual-drive LiNbO3 (LN) Mach-Zehnder modulator (MZM) with a Vπ of 2.2 V. The optical comb signal was then passed through a programmable pulse shaper  followed by a single-mode fiber (SMF) for chirp compensation, where a 10 GHz coherent Gaussian pulse was generated. The gray and red curves in Fig. 2(a) show the optical comb and shaped pulse spectra, respectively. The autocorrelation trace of the 10 GHz pulse is shown in Fig. 2(b). The spectral width and pulse duration were 1.5 nm and 2.4 ps, respectively. The time-bandwidth product was 0.45, which indicated that a transform-limited Gaussian pulse was obtained.
A 10 GHz, 64 QAM coherent optical pulse signal was generated by passing the pulse train through an IQ modulator, driven with a 10 GHz, 64 QAM base-band signal from an arbitrary waveform generator (AWG). Here, we pre-compensated for the nonlinear phase rotation induced by self phase modulation (SPM) that occurs during transmission. The RZ-QAM signal was then passed through an OTDM multiplexing circuit where the symbol rate was increased to 160 Gsymbol/s. After that, the data signal was polarization multiplexed by using a polarization beam combiner. Simultaneously, the 28th harmonic of the optical comb signalwas extracted by a 6.5 GHz optical filter. This was used as a pilot tone signal for OPLL at the receiver. The combined data and pilot signals were fed into a transmission link.
Figure 3(a) shows the optical spectrum of the 1.92 Tbit/s, 64 QAM data and pilot signals. The optical bandwidth at −20 dB including the pilot tone signal was 4.5 nm (562.5 GHz). The corresponding time waveform is also shown in Fig. 3(b). The transmission link was a dispersion-managed fiber consisting of two 75 km spans composed of a 50 km super large area (SLA) fiber with a dispersion of 19.5 ps/nm/km and a 25 km inverse dispersion fiber (IDF) with a −40 ps/nm/km dispersion. The average loss was 18 dB/span, which was compensated for with EDFAs. Figure 4 shows the bit error rate (BER) of the demodulated 10 Gsymbol/s, 64 QAM signal after 150 km transmission for various powers launched into each span, which was measured after undergoing OTDM demultiplexing, RZ-CW conversion and coherent homodyne detection whose details are described below. From these results, the launch power was set at the optimal value of 4 dBm. Figure 5 shows the optical spectra of the data signal before and after a 150 km transmission at the 2 nm resolution bandwidth. The optical signal-to-noise ratios (OSNRs) of these signals were 37 and 28 dB, respectively, which corresponded to respective OSNRs of 50 and 41 dB when measured under 0.1 nm resolution. Hence the OSNR of the data signal degraded by 9 dB during transmission.
After transmission, the 1.92 Tbit/s data signal was polarization-demultiplexed with a polarization beam splitter. Subsequently, OTDM demultiplexing was achieved using a nonlinear optical loop mirror (NOLM). The NOLM consisted of a 100 m highly nonlinear fiber with γ = 20.4 W−1km−1, a dispersion slope of 0.029 ps/nm2/km, and a zero dispersion wavelength of 1522 nm. Here, a control pulse was generated by using a 1564.7 nm CW DFB-LD followed by an optical comb generator , which was driven by a clock signal extracted from the QAM data signal , and an optical filter. Figures 6(a) and 6(b) show the optical spectrum and time waveform of the control pulse. The spectral width and pulse duration were 0.8 nm and 5.4 ps, respectively.
The 10 Gsymbol/s, 64 QAM data signal obtained after the demultiplexing was then passed through an RZ-CW conversion circuit composed of a dispersion compensation fiber (DCF) with a dispersion of −69 ps/nm and an LN phase modulator driven by an extracted clock signal with a modulation depth of 2.5π . RZ-CW conversion results in the narrowing of the spectral width and the increase in the peak power at the central frequency. This enables demodulation of data signals at a higher signal to noise ratio (SNR) within the demodulation bandwidth. A CW-local oscillator (LO), a fiber ring laser with a linewidth of 4 kHz , was used for the homodyne detection of the RZ-CW converted data signal. This CW-LO was phase-locked to the data signal using an OPLL as illustrated in Fig. 7, which shows the frequency relationship between the optical comb spectrum at the transmitter and the phase-modulated CW-LO signal at the OPLL circuit. In the OPLL circuit, a CW-LO signal was phase-modulated at 4fclock-100 MHz. The phase of the beat signal between the 7th harmonic of the modulated CW-LO signal, whose frequency was shifted by 7 × (4fclock-100 MHz) from the LO frequency, and the transmitted pilot tone signal was compared with the reference phase from the synthesizer at an offset frequency of 700 MHz by the double balanced mixer (DBM). The DBM output the voltage phase error signal between these signals. This signal was then fed back to the LN phase modulator in the CW-LO cavity so as to control the CW-LO phase. Thus, the CW-LO frequency was phase-locked to the transmitter frequency, and homodyne detection could be easily realized between an RZ signal and a synchronized CW-LO signal. Figures 8(a) and 8(b) show the 700 MHz-intermediate frequency (IF) signal spectrum measured with an electrical spectrum analyzer with a span of 2 MHz and the single side band (SSB) phase noise spectrum. The phase noise was 1.8 deg., which was less than the phase allowance for the 64 QAM of 4.7 deg. This phase allowance is the smallest phase difference between two adjacent symbols in the 64 QAM constellation map.
After homodyne detection, the received signal was A/D converted at a sampling rate of 40 Gsample/s and demodulated in an offline condition with a digital signal processor (DSP). Here, the demodulation bandwidth was set at 10 GHz with a digital filter in the DSP. At the DSP, we compensated for the waveform distortion by using FDE. In our FDE process, a fixed pattern 64 QAM signal was used as a training sequence. During transmission, this training sequence gets distorted due to the non-ideal frequency response of individual components such as IQ modulators. The transmitted training symbol Htrans(ω) was then compared with the non-distorted training symbol Hideal(ω) in the frequency domain, where we obtained a distortion function Fdist(ω) = Htrans(ω)/Hideal(ω). As a result, we were able to compensate for waveform distortion to the data signal by dividing the spectrum by the distortion function. The FFT size was 16384, which resulted in a frequency resolution of 2.44 MHz. This FFT size is the minimum block size for which we obtained the best error vector magnitude (EVM) of the demodulated data signal. The bit error rate (BER) was calculated after the data had been converted back to the time domain by IFFT. For comparison, we also separately used a 99-tap FIR filter with a frequency resolution of 100 MHz, which was the minimum resolution in our previous demodulation circuit . We set the number of FIR filter taps at 99 to avoid the slow convergence of the tap coefficient calculation with the least-mean-square (LMS) algorithm for larger number of taps.
Here, we show the advantage of FDE as regards the computational complexity as a function of frequency resolution. The frequency resolution of FDE and FIR filter is expressed as ΔfFDE = sampling rate/NFFT and ΔfFIR = symbol rate/NFIR, respectively. Here, NFFT is the FFT size for FDE and NFIR is the number of FIR filter taps. The computational complexity, which is defined as the number of real-valued multiplications per symbol, is given by nFDE = 8log2(NFFT) for FDE and nFIR = 4NFIR for FIR filter . Figure 9 shows the relationship between the number of real-valued multiplications and the frequency resolution. From this figure, it can be clearly seen that FDE requires a much less number of multiplications than FIR filter especially for higher resolutions.
3. Experimental results
First, we show the improved waveform distortion compensation performance using FDE. Figures 10(a) and 10(b) show the back-to-back constellation of a demodulated 10 Gsymbol/s, 64 QAM signal using an FIR filter and FDE, respectively. With the adoption of FDE, the EVM decreased from 4.4 to 3.6%. This improvement is a consequence of the ability of FDE to compensate for waveform distortions with a high resolution better than that of an FIR filter.
Figure 11 shows the optical spectra of the 10 Gsymbol/s, 64 QAM signal before and after RZ-CW conversion. The RZ-CW conversion process enabled the spectral width to be reduced, and the OSNR at the central frequency to be increased by as much as 5 dB.
Figure 12(a) shows the BER characteristics as a function of the received power for one tributary under back-to-back conditions and after a 150 km transmission. The BERs for all the tributaries after a 150 km transmission at a received power of −16 dBm are shown in Fig. 12(b). After a 150 km transmission, there was a power penalty of 5 dB at a BER of 2 × 10−3. For all 16 tributaries, BERs were obtained that were below the forward error correction (FEC) limit of 2 × 10−3. The power penalty observed after transmission is a result of OSNR degradation during transmission. As shown in Fig. 12(a), when we carried out a transmission with a single polarization, there was no error floor in the BER characteristics. Here, the launch power was set at an optimum value of 1.5 dBm. However, with pol-mux, there was an error floor in the BER characteristics. This is therefore mainly attributed to cross phase modulation (XPM) between the two polarizations. This transmission is scalable to a net spectral efficiency of 3.2 bit/s/Hz, considering an optical bandwidth of 562.5 GHz at −20 dB and a 7% FEC overhead.
We successfully demonstrated a single-channel, 1.92 Tbit/s, 64 QAM coherent optical pulse OTDM transmission over 150 km. This is the highest QAM multiplicity yet employed in a coherent pulse OTDM transmission. We were able to obtain these results by combining RZ-CW conversion and FDE techniques. In this transmission, we can achieve a spectral efficiency of 3.2 bit/s/Hz in a multi-channel system when we take the 7% FEC overhead into account.
We thank T. Hara and S. Oikawa of Sumitomo Osaka Cement Co., Ltd. for providing a low Vπ, dual-drive LN Mach-Zehnder modulator for optical comb generation.
References and links
1. P. Winzer, “Beyond 100G ethernet,” IEEE Commun. Mag. 48(7), 26–30 (2010). [CrossRef]
2. M. Nakazawa, T. Yamamoto, and K. R. Tamura, “1.28 Tbit/s-70 km OTDM transmission using third- and fourth-order simultaneous dispersion compensation with a phase modulator,” Electron. Lett. 36(24), 2027–2029 (2000). [CrossRef]
3. P. Guan, T. Hirano, K. Harako, Y. Tomiyama, T. Hirooka, and M. Nakazawa, “2.56 Tbit/s/ch Polarization-multiplexed DQPSK transmission over 300 km using time-domain optical Fourier transformation,” Opt. Express 19(26), B567–B573 (2011). [CrossRef] [PubMed]
4. E. Palushani, C. Schmidt-Langhorst, T. Richter, M. Nolle, R. Ludwig, and C. Schubert, “Transmission of a serial 5.1-Tb/s data signal using 16-QAM and coherent detection,” in Proceedings of the Euro.Conf. on Optical Communication (ECOC), Geneva, 2011, We.8.B.5. [CrossRef]
5. T. Richter, C. Schmidt-Langhorst, M. Nolle, R. Ludwig, and C. Schubert, “Single wavelength channel 10.2 Tb/s TDM-capacity using 16-QAM and coherent detection,” in Proceedings of the Optical Fiber Communication Conference (OFC), Los Angeles, 2011, PDPA9.
6. M. Nakazawa, K. Kasai, M. Yoshida, and T. Hirooka, “Novel RZ-CW conversion scheme for ultra multi-level, high-speed coherent OTDM transmission,” Opt. Express 19(26), B574–B580 (2011). [CrossRef] [PubMed]
7. K. Kasai, D. O. Otuya, M. Yoshida, T. Hirooka, and M. Nakazawa, “Single-carrier 800-Gb/s 32 RZ/QAM coherent transmission over 225 km employing a novel RZ-CW conversion technique,” IEEE Photon. Technol. Lett. 24(5), 416–418 (2012). [CrossRef]
8. K. Ishihara, T. Kobayashi, R. Kudo, Y. Takatori, A. Sano, E. Yamada, H. Masuda, and Y. Miyamoto, “Coherent optical transmission with frequency-domain equalization,” in Proceedings of the Euro.Conf. on Optical Communication (ECOC), Brussels, 2008, We.2.E.3.
9. Y. Koizumi, K. Toyoda, T. Omiya, M. Yoshida, T. Hirooka, and M. Nakazawa, “512 QAM transmission over 240 km using frequency-domain equalization in a digital coherent receiver,” Opt. Express 20(21), 23383–23389 (2012). [CrossRef] [PubMed]
10. D. O. Odeke, K. Kasai, T. Hirooka, M. Yoshida, M. Nakazawa, T. Hara, and S. Oikawa, “A single-channel, 1.6 Tbit/s 32 QAM coherent pulse transmission over 150 km with RZ-CW conversion and FDE technique,” in Proceedings of the Optical Fiber Communication Conference (OFC), Anaheim, 2013, OTh4E.4.
11. K. Kasai, A. Suzuki, M. Yoshida, and M. Nakazawa, “Performance improvement of an acetylene (C2H2) frequency-stabilized fiber laser,” IEICE Electron. Express 3(22), 487–492 (2006). [CrossRef]
12. T. Sakamoto, T. Kawanishi, and M. Izutsu, “Asymptotic formalism for ultraflat optical frequency comb generation using a Mach-Zehnder modulator,” Opt. Lett. 32(11), 1515–1517 (2007). [CrossRef] [PubMed]
13. G. Baxter, S. Frisken, D. Abakoumov, H. Zhou, I. Clarke, A. Bartos, and S. Poole, “Highly programmable wavelength selective switch based on liquid crystal on silicon switching elements,”inProceedings of the Optical Fiber Communication Conference (OFC),Anaheim, 2006, OTuF2. [CrossRef]
14. C. Boerner, V. Marembert, S. Ferber, C. Schubert, C. Schmidt-Langhorst, R. Ludwig, and H. G. Weber, “320 Gbit/s clock recovery with electro-optical PLL using a bidirectionally operated electroabsorption modulator as phase comparator,” in Proc. of the Optical Fiber Communication Conf. (OFC), Anaheim, 2005, OTuO3. [CrossRef]
15. K. Kasai, J. Hongo, M. Yoshida, and M. Nakazawa, “Optical phase-locked loop for coherent transmission over 500 km using heterodyne detection with fiber lasers,” IEICE Electron. Express 4(3), 77–81 (2007). [CrossRef]