We describe a single-channel 10.2 Tbit/s online transmission using non-coherent ultrashort optical Nyquist pulses. A 10.2 Tbit/s signal was generated at a symbol rate of as fast as 2.56 Tbaud with a polarization-multiplexed DQPSK format. We developed a new ultrafast optical sampler for Nyquist OTDM demultiplexing with a nonlinear optical loop mirror, an RZ-CW conversion technique to improve the SNR, and an active stabilization technique providing stable long-term demultiplexing operation. With precise higher-order dispersion compensation up to fourth order, a 10.2 Tbit/s signal was transmitted over 300 km for the first time as a real-time demonstration with a spectral efficiency of 2.5 bit/s/Hz. We also report a 10.2 Tbit/s transmission over 225 km with a spectral efficiency of 3.7 bit/s/Hz, which we realized by reducing the roll-off factor to zero.
© 2018 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
The rapid growth of cloud computing and data center traffic has driven intensive research towards faster network interfaces with a single-channel bit rate of 1 Tbit/s and beyond. Recent advances in high-speed digital and analog electric devices such as digital-to-analog converters and multiplexers have made it possible to realize symbol rates exceeding 100 Gbaud. This includes recent demonstrations of 180 Gbaud electrical time division multiplexed (ETDM) polarization-division multiplexed quadrature phase shift keying (PDM-QPSK)  and 100 Gbaud PDM-64 quadrature amplitude modulation (QAM) transmissions , which correspond to line rates of 720 Gbit/s and 1.2 Tbit/s, respectively. A 1-Tbit/s PDM-16 QAM signal has also been generated at 125 Gbaud by combining × 2 ETDM and × 2 optical time division multiplexing (OTDM) from a 31.25 Gbaud signal . However, the symbol rates are limited by the speed and bandwidth of electrical and electro-optical components, and therefore faster symbol rates such as 1 Tbaud have only been realized with OTDM [4,5]. For example, there has been a preliminary report of a 10.2-Tbit/s PDM-16 QAM transmission over 29 km at a symbol rate of 1.28 Tbaud . However, it was only realized with a self-homodyne detection scheme where a local oscillator (LO) pulse is supplied directly from the transmitter. In addition, despite the adoption of 16 QAM, the spectral efficiency (SE) was limited to approximately 2 bit/s/Hz since the experiment employed spectrally inefficient Gaussian pulses.
As a spectrally efficient optical pulse suitable for ultrahigh-speed OTDM, we have proposed an optical Nyquist pulse and its orthogonal TDM scheme . A Nyquist pulse has a narrow spectral profile, which is typically less than half that of a Gaussian pulse, and this makes it advantageous in terms of both its high SE  and greater tolerance to chromatic dispersion (CD) and polarization-mode dispersion (PMD) . By taking advantage of these features, we have demonstrated a 7.68 Tbit/s, PDM-64 QAM transmission at 640 Gbaud with an SE of 9.7 bit/s/Hz . We previously reported a 5.12 Tbit/s PDM-differential quadrature phase shift keying (DQPSK) transmission over 300 km at a symbol rate of 1.28 Tbaud, using a 680 fs non-coherent Nyquist pulse .
In this paper, by reducing the pulse width of a non-coherent Nyquist pulse to 340 fs and increasing the symbol rate from 1.28 to 2.56 Tbaud, we demonstrate a single-channel 10.2 Tbit/s online transmission over a record distance of 300 km. It is very important to investigate scientifically the feasibility of ultrafast single-channel transmission when using femtosecond optical pulse technology, which may pave the way toward ultrafast TWDM transmission in the near future. To achieve this 10 Tbit/s transmission, we developed a new technique for precise higher-order dispersion compensation up to the fourth order with a programmable optical filter, a nonlinear optical loop mirror (NOLM) newly developed for 2.56 Tbaud to 40 Gbaud demultiplexing, a return-to-zero (RZ)-continuous wave (CW) conversion technique for improving the signal-to-noise ratio (SNR) , and an active timing stabilization technique that provides stable long-term demultiplexing operation. As a result, a 10.2 Tbit/s PDM-DQPSK signal was transmitted over 300 km with an SE of 2.5 bit/s/Hz using a Nyquist pulse with a roll-off factor of α = 0.5. We also demonstrate a 10.2 Tbit/s transmission over 225 km using a Nyquist pulse with α = 0, in which the SE increased to 3.7 bit/s/Hz.
2. Experimental setup for single-channel 10.2 Tbit/s-300 km Nyquist pulse transmission
Figure 1 shows our experimental setup for a single-channel 10.2 Tbit/s-300 km transmission. Although the setup is based on that reported for a 5.12 Tbit/s transmission , a much more precise handling of the ultrafast Nyquist pulses is required throughout the system in a 10.2 Tbit/s transmission as we describe below.
2.1 10.2 Tbit/s Nyquist pulse transmitter
As a transmitter pulse source, we used a 40 GHz mode-locked fiber laser (MLFL) emitting a 1.4 ps Gaussian pulse at 1543 nm. To generate an ultrashort Nyquist pulse for 2.56 Tbaud transmission with a spectral width of 30 nm (3.84 THz) for a roll-off factor α = 0.5, the spectrum was broadened over almost the entire C-band in a highly nonlinear dispersion-flattened fiber (HNL-DFF) with a normal dispersion at a launch power of 24 dBm (see the dashed-line box (a) in Fig. 1). It was then DQPSK modulated at 40 Gbaud with a 211–1 PRBS, whose length was limited due to DQPSK decoding. The DQPSK signal was multiplexed to 2.56 Tbaud with OTDM bit interleavers, which is emulated by delay-line Mach-Zehnder interferometers (MZI) on a planar lightwave circuit (PLC). Each multiplexer is composed of three multiplexing stages, namely three cascaded MZIs with different delays, where the first circuit is designed for 40 → 320 Gbaud OTDM and the second for 320 Gbaud → 2.56 Tbaud OTDM (see the dashed-line box (b) in Fig. 1). Their insertion loss is 13 dB, including a 9 dB inherent loss due to 3 dB couplers at three multiplexing stages. An EDFA is inserted between the two multiplexers to compensate for the loss. The arms of the MZIs are equipped with heaters to allow the precise tuning of the optical path length. At each MZI stage, the pulses are interleaved by 1.5 times the pulse interval, where the magnitude of the interleaving is limited by the physical dimensions of the PLC. The OTDM signal thus obtained is then shaped into optical Nyquist pulses by using a programmable optical filter as a pulse shaper.
The optical spectra at the transmitter are shown in Fig. 2(a)-2(c). Figure 2(a) shows the optical spectrum at the output of the MLFL (black curve) and that after spectral broadening and DQPSK modulation (red curve). The spectrum is shaped into that of a Nyquist pulse with α = 0.5 by using a transfer function plotted in Fig. 2(b). The obtained spectrum of the Nyquist pulse generated at 40 GHz is shown by the red curve in Fig. 2(c), where the OTDM bit interleavers are replaced by optical attenuators with an identical loss. The obtained spectrum accurately fits the ideal Nyquist pulse spectrum shown by the blue curve. Figure 2(d) shows an autocorrelation waveform of the Nyquist pulse generated at 40 GHz. As a reference, the black curves show the actual waveform of a Nyquist pulse and an autocorrelation waveform that was obtained theoretically. The zero-crossing period was as short as 390 fs, which corresponds to a symbol rate of 2.56 Tbaud after OTDM. The periodic zero crossing cannot be observed in the autocorrelation waveform, because two Nyquist pulses separated by the zero-crossing period still overlap, and hence their overlap integral does not become zero. Instead, the auto-correlation waveform exhibits small changes in the curvature of the tail due to the zero-crossing property in the actual waveform. From the autocorrelation measurement, the FWHM is estimated to be 340 fs, which agrees well with the ideal FWHM of 330 fs.
Figure 3 shows the waveform of a generated Nyquist OTDM signal that was measured using an optical sampling oscilloscope with a resolution of 500 fs. Figure 3(a) shows a 1.28 Tbaud signal where every other pulse is omitted from a 2.56 Tbaud signal. Individual pulses can be clearly identified. In the 2.56 Tbaud signal shown in Fig. 3(b), it is difficult to identify the symbol period due to the overlapped TDM of the Nyquist pulses.
2.2 300 km transmission line
After polarization multiplexing, the 10.2 Tbit/s data were launched into a 300 km transmission link consisting of 4 × 75 km spans. Each span was composed of a 50 km single-mode fiber (dispersion: 20 ps/nm/km, dispersion slope: 0.07 ps/nm2/km) and a 25 km inverse dispersion fiber (dispersion: −40 ps/nm/km, dispersion slope: −0.14 ps/nm2/km) so that the second- and third-order dispersions were compensated for simultaneously. The loss of each span was compensated for with an EDFA and Raman amplifiers. The Raman amplifiers were backward pumped and provided a 10-dB gain with a pump power of 500 mW, while the EDFA gain was 6 dB. The gain flatness of the in-line EDFA and Raman amplifiers was < 1 dB, and this played an important role in maintaining the flat-top spectrum of the Nyquist pulse. The first-order PMD was mitigated by coupling the transmission signal along the principal state of polarization (PSP) of the 300 km transmission link using polarization controllers. The residual second-order dispersion was precisely compensated for after the 300 km transmission by using a grating-pair type variable dispersion compensator with an accuracy of < 0.001 ps/nm. As regards the higher-order dispersion, both the residual third-order dispersion, as in the 5.12 Tbit/s transmission , and the fourth-order dispersion, were precisely compensated for with spectral phase manipulation by applying a phase profile ϕ(ω) = exp(−iβ3Lω3/6 − iβ4Lω4/24) at the pulse shaper before transmission with accuracies of |β3L| < 0.01 ps3 and |β4L| < 0.001 ps4, where β3 and β4 are the third- and fourth-order dispersion coefficients, respectively, and L ( = 300 km) is the length. The waveforms of a Nyquist OTDM signal after transmission over 300 km, measured for 1.28 and 2.56 Tbaud, are shown in Fig. 4(a) and 4(b), respectively. Compared with Fig. 3(a), although the OSNR is degraded, individual pulses can still be clearly identified after the 300 km transmission in Fig. 4(a) thanks to the accurate compensation of CD up to the fourth order.
2.3 2.56 Tbaud → 40 Gbaud demultiplexer using ultrafast NOLM
After the transmission, the signal was first separated into two orthogonal polarization channels using a polarization-beam splitter (PBS) and then demultiplexed from 2.56 Tbaud to 40 Gbaud. To extract the ideal ISI-free points existing only at the symbol period from the interleaved Nyquist pulses, we developed a new optical sampling method that is faster than the symbol period (390 fs) by using a NOLM. As a NOLM control pulse source, we used a 40 GHz MLFL operating at 1593 nm and emitting a 760 fs pulse . The repetition rate was synchronized to the transmitted data through a phase-locked loop (PLL) with an extracted 40 GHz clock, and an electro-optical PLL circuit with an electro-absorption modulator used as a phase comparator  was adopted for the clock extraction. Figure 5(a) shows the RF spectrum of a 40 GHz clock signal extracted from 2.56 Tbaud data after a 300 km transmission, and Fig. 5(b) shows the single sideband (SSB) phase noise power density vs. the offset frequency at the 40 GHz carrier frequency. By integrating the SSB phase noise spectrum from 10 Hz to 1 MHz, the jitter was estimated to be as small as 57 fs.
To obtain an ultrashort control pulse for the NOLM, the MLFL output pulse was externally compressed in a similar way to the transmitter, namely by broadening the spectrum with HNL-DFF with a launch power of 28 dBm, followed by chirp compensation and spectral shaping with a pulse shaper. Here, the control pulse was shaped into a 280 fs sech pulse, whose autocorrelation waveform and optical spectrum are shown in Fig. 6. A sech control pulse is more beneficial than a Gaussian pulse as it can propagate inside the NOLM with minimum distortion of the waveform and spectrum due to the soliton effect. This even makes it possible to compress the control pulse during propagation. These features offer a substantial advantage in terms of ultrafast NOLM operation as we describe below.
The NOLM is newly designed for ultrafast optical sampling for a 2.56 Tbaud signal. It is composed of a 20 m highly nonlinear fiber (HNLF) with a nonlinear coefficient of γ = 22 W−1 km−1, a zero-dispersion wavelength of 1568 nm, and a dispersion slope of 0.013 ps/nm2/km. Such a low dispersion slope plays an important role in reducing the dispersion at the signal wavelength (−6.5 × 10−3 ps/nm at λs = 1543 nm) so that signal pulse broadening in the NOLM can be alleviated. In addition, since the wavelength of the control pulse (λc = 1593 nm) is located in the anomalous dispersion regime (6.5 × 10−3 ps/nm), it is possible to incorporate the soliton effect as we described earlier. A walk-off-free operation was realized with the optimum and precise wavelength setting of the signal and control pulses. The NOLM also includes polarization controllers for optimizing the state of polarization of both the signal and control pulses to maximize the extinction ratio and switching efficiency. Here, two WDM couplers were employed; one to couple the control pulse into the loop and the other to separate the signal and control pulses at the output. The insertion loss of the NOLM including the isolators, polarization controllers and WDM couplers was 5 dB, and the maximum extinction ratio reached as high as 36 dB.
To evaluate the performance of the fabricated NOLM, we first measured the sampling gate width by launching a CW probe light into the NOLM instead of an OTDM signal. Here, the control pulse was coupled into the NOLM through a WDM coupler at an input power of 18 dBm. An autocorrelation waveform of the sampling gate thus obtained is shown in Fig. 7. The sampling gate width was as short as 230 fs, and the waveform accurately fits a sech profile. We then launched a 2.56 Tbaud Nyquist OTDM signal at an input power of 27 dBm and demultiplexed it to 40 Gbaud using the NOLM. The demultiplexed waveform is shown in Fig. 8(a). One tributary was selected for measurement with an optical delay line in front of the clock extraction circuit with fine-tuning capability. It can be seen that a tributary is extracted without leakage from other tributaries. Figure 8(b) shows the spectrum of the demultiplexed data and control pulses measured at the NOLM output. The control pulse spectrum consists of a number of neatly repetitive spectral lines, corresponding to the 40 GHz-spaced longitudinal modes of the pulse train as shown in Fig. 6(b), while the demultiplexed data pulse has a continuous spectrum. The spectra of the data and the control pulses were sufficiently separated during OTDM demultiplexing.
The NOLM gate width was optimized by varying the pulse width of the control pulse. Figure 9 shows the relationship between the gate width and the back-to-back bit error rate (BER). This indicates that the optimum gate width is 230 fs, which is determined by the trade-off between the greater penetration from adjacent tributaries caused by a broader gate width and the decrease in OSNR after demultiplexing caused by a narrower gate width.
2.4 NOLM timing stabilization
Because of such ultrafast sampling, the demultiplexing is very sensitive to the timing drift between the signal and control pulses. Although the repetition rate of the control pulse source is synchronized with the transmitted data signal through the PLL operation of the MLFL, timing drift occurs in the optical path between the MLFL output and the NOLM gate input, which causes an unstable demultiplexing operation. In particular, the optical path includes a 500 m HNL-DFF, two EDFAs and a pulse shaper, which are all indispensable for compressing the control pulse for ultrafast demultiplexing but they introduce timing drift. To reduce this drift, the phase of a 40 GHz clock output from the MLFL and that extracted just before the NOLM are compared with a double-balanced mixer (DBM), and the phase difference converted into an error signal is fed back to a piezo electric transducer (PZT) on which a part of the fiber is wound (see dashed-line box (c) in Fig. 1). Here, a proportional and integral (PI) control circuit is employed. Figure 10 shows the temporal profile of a change in the detected phase difference that is converted to a voltage signal in a free-running condition and under stabilization. Without stabilization, drift is observed with a scale of one symbol period within one minute. With stabilization, the drift is controlled to within 40 fs, which makes it possible to realize a stable long-term demultiplexing operation.
2.5 RZ-CW converter
After the OTDM demultiplexing, the 40 Gbaud data were separated from the control pulses with a WDM coupler and passed through an RZ-CW conversion circuit (see dashed-line box (d) in Fig. 1) for spectral compression . It should be noted that the pulse width of a demultiplexed 40 Gbaud signal is extremely short, and therefore the spectrum is excessively broad. Such a broad spectrum is disadvantageous in terms of O/E conversion efficiency at a photo detector due to its limited bandwidth and inevitably leads to the detection of a low SNR. Low-pass filtering of the broad spectrum also leads to OSNR degradation due to excessive filtering loss. On the other hand, in principle RZ-CW conversion enables the spectrum to be narrowed without sacrificing the OSNR. The RZ-CW converter consists of a dispersion-compensating fiber (DCF) as a dispersive element and an LN phase modulator for linear chirping. In the circuit, the 40 Gbaud signal is first broadened with DCF with a group-velocity dispersion (GVD) of −1.0 ps/nm, after which it is given an opposite chirp using the phase modulator driven by the 40 GHz clock with a modulation depth of 5π. The resultant chirp-compensated signal has a narrower spectrum than an input pulse. If the pulse is spread over a sufficiently broad width and a strong linear chirp is applied, the data encoded on an RZ pulse can be ideally converted to the data encoded on a CW carrier.
Figure 11 shows the result of RZ-CW conversion applied to a demultiplexed 40 Gbaud signal, where the waveform measured with an optical sampling oscilloscope and the optical spectrum before and after RZ-CW conversion are plotted in Fig. 11(a) and 11(b), respectively. Figure 11(b) shows that a 5 dB SNR improvement at the spectral peak is obtained as a result of the spectral narrowing. This yields a BER improvement of a half to one order of magnitude. Here, the conversion efficiency is determined by the amount of GVD that is applied to avoid any overlap with adjacent pulses, linearity in the frequency chirp of the sinusoidal driving voltage, and the maximum chirp limited by the driver voltage.
The 40 Gbaud DQPSK signal after the RZ-CW conversion was then demodulated with a one-bit delay interferometer and detected with a 40 GHz balanced photo diode (PD). Finally, a BER measurement was carried out online with an error detector.
3. Experimental results for 10.2 Tbit/s-300 km Nyquist pulse transmission
We first optimized the transmission power by measuring the BER of one specific tributary after 300 km transmissions at different launch powers. Figure 12 shows the launch power dependence of the BER in a 10.2 Tbit/s-300 km transmission. Based on this result, we set the fiber launch power at 11 dBm with a view to maximizing the OSNR and minimizing nonlinear impairments. The optical spectra of a 10.2 Tbit/s signal before and after a 300 km transmission obtained for this launch power is shown in Fig. 13. Here, eight sharp peaks can be seen in the optical spectrum of the OTDM signal, whose envelope can be fitted with the spectrum of a Nyquist pulse of α = 0.5 as shown by the blue dotted curve. These peaks are attributed to the phase relationship of the bit interleaved pulses at the OTDM emulator. Specifically, at each multiplexing stage, the pulses are interleaved by only 1.5 times the symbol interval as we described in Sec. 2.1, which results in a residual correlation in the data sequence of the OTDM signal. The interval between the peaks is 427 GHz, which corresponds to (1/6) × 2.56 THz, indicating that a correlation exists at every six symbols in the 2.56 Tbaud OTDM signal. These spectral modifications make it difficult to define OSNR exactly, but assuming that the ratio between the peak and the ASE noise level provides a rough approximation of the OSNR, the measured values were 46 and 38 dB before and after a 300 km transmission, respectively, and the noise level rose by 8 dB.
Figure 14 shows the waveform of a 40 Gbaud DQPSK pulse demultiplexed from the transmitted 2.56 Tbaud signal and its demodulated waveform with a balanced PD. Here, the result for a single polarization transmission is shown in Fig. 14(a) while Fig. 14(b) is the result for a polarization-multiplexed transmission. A comparison of Fig. 14(a) and 14(b) shows that polarization multiplexing causes a large impairment where the pulse peak fluctuates more severely. This is mainly attributed to polarization crosstalk induced by second-order PMD (depolarization), which inevitably occurs due to the broad spectral width even if the first-order PMD is mitigated . To evaluate the significance of the polarization crosstalk, we measured the optical spectrum of one polarization channel and that of the component leaked to the other polarization channel after a 300 km transmission. The results of measurements at each output port of PBS are shown in Fig. 15(a) by the red and blue curves, respectively. It can be seen that the signal suffers greatly from crosstalk, which had an approximate measured value of –16 dB. We also measured the polarization crosstalk for various transmission distances, and the result is shown in Fig. 15(b). The result for a 1.28 Tbaud Nyquist pulse is also shown for comparison. The crosstalk grows by the square of the distance, which has also been analytically verified . It can also be observed that the growth is very rapid at 2.56 Tbaud compared with the 1.28 Tbaud transmission.
Finally, we carried out an online BER measurement of a 10.2 Tbit/s-300 km transmission for all tributaries. Figure 16 shows the relationship between the BER and the received power measured for one tributary (Y polarization of #7 in Fig. 17). Here, the received power is defined as Prec in Fig. 2, namely it is the input optical power to the preamplifier, which was varied with an optical attenuator. In a back-to-back configuration, error-free performance was obtained at a received power of – 25 dBm. After a 300 km transmission, the BER curve had an error floor at 2 × 10−6 in a single-polarization transmission and at 1.5 × 10−4 in a polarization-multiplexed transmission. The BERs for all 64 tributaries in the single-polarization and polarization-multiplexed transmissions are plotted in Figs. 17(a) and 17(b), respectively. As seen in these figures, the BER is degraded by approximately two orders of magnitude with polarization multiplexing. This BER degradation is attributed to polarization crosstalk caused by second-order PMD (depolarization) as described above. Despite this impairment, a BER below the forward-error correction (FEC) threshold of 2 × 10−3 (7% overhead) was achieved for all tributaries. As a result, a 10.2 Tbit/s signal (net data rate of 9.5 Tbit/s) was transmitted over 300 km in a 3.84 THz bandwidth, which corresponds to an SE of 2.5 bit/s/Hz.
4. 10.2 Tbit/s-225 km transmission with an SE of 3.7 bit/s/Hz using α = 0 Nyquist pulse
The experiments described above were realized using a Nyquist pulse with a roll-off factor α = 0.5, where the spectral width was 30 nm (3.84 THz) and the corresponding SE was 2.5 bit/s/Hz. An important feature of the Nyquist pulse is that the spectral width can be varied by changing the α value, which enables the SE to be managed. By reducing the spectral width with a lower α value, it is possible to improve the SE. Here, we present a 10.2 Tbit/s transmission experiment using a Nyquist pulse with α = 0, in which the spectral width is reduced to 20 nm (2.56 THz) and correspondingly the SE is improved to 3.7 bit/s/Hz. No other pulses are capable of this spectral manipulation.
The experimental setup follows that shown in Fig. 1, and we only changed the transfer function at the pulse shaper to generate a Nyquist pulse with α = 0. The newly designed transfer function is shown in Fig. 18(a), and the optical spectrum and autocorrelation waveform of the generated Nyquist pulse at 40 GHz are shown in Fig. 18(b) and 18(c), respectively. The actual waveform of the Nyquist pulse and the autocorrelation waveform obtained theoretically are shown by the black curves in Fig. 18(c). Here, the auto-correlation waveform for α = 0 can be analytically obtained asFig. 2(d), periodic zero crossing cannot be observed in the autocorrelation waveform as can also be understood from this equation. From the autocorrelation measurement, the FWHM is estimated to be 342 fs, which agrees well with the ideal FWHM (345 fs) for α = 0. The optical spectrum of a 10.2 Tbit/s signal is shown in Fig. 18(d). The envelope of six sharp peaks can be fitted with the spectrum of a Nyquist pulse of α = 0 as shown by the blue dotted curve.
Figure 19(a) shows the waveform of a 40 Gbaud signal demultiplexed from a 2.56 Tbaud Nyquist OTDM signal with α = 0, which was measured in a back-to-back configuration. Figure 19(b) shows the spectrum of the OTDM data and control pulses measured at the NOLM output. Here, the sampling gate width was set at 230 fs as with α = 0.5, and the power of the control pulse coupled into the NOLM was set at 16 dBm. This power level was 2 dB less than for α = 0.5. We reduced the power of the control pulse to avoid any spectral overlap with the Nyquist pulse, whose flat-top spectral width becomes broader for α = 0 than with α = 0.5. Figure 19(a) shows that a data signal is successfully demultiplexed, although the fluctuation in the peak intensity is slightly increased compared with that for α = 0.5 shown in Fig. 8(a). Since the ringing in the tail of an α = 0 Nyquist pulse oscillates over a longer time than that for α = 0.5, the ISI from other time slots becomes larger.
Figure 20 shows the relationship between the BER and the received power in a 10.2 Tbit/s-225 km transmission with α = 0, measured for one tributary (Y polarization of #7 in Fig. 21). In a back-to-back configuration, error-free performance was obtained at a received power of > – 24 dBm. The receiver sensitivity is 1 dB lower than with α = 0.5, which may be attributed to the greater penetration of the Nyquist pulse tails for a lower α value. After a 225 km transmission, the BER curve had an error floor at 1 × 10−7 in a single-polarization transmission and at 3 × 10−5 in a polarization-multiplexed transmission. The BERs for all 64 tributaries are plotted in Fig. 21. The worst BER was 1.3 × 10−3, which means that a BER below the FEC threshold of 2 × 10−3 was obtained for all tributaries. This is the first demonstration of a single-channel 10 Tbit/s transmission with an SE as high as 3.7 bit/s/Hz achieved with a non-coherent Nyquist pulse.
We demonstrated the first single-channel 10.2 Tbit/s transmission of PDM-DQPSK signals over 300 km using ultrashort optical Nyquist pulses, where online demodulation was realized with non-coherent detection. The symbol rate was increased to 2.56 Tbaud using a 340-fs ultrashort Nyquist pulse with a 30 nm bandwidth, which was realized by taking full advantage of the entire C-band. The spectral width was then further decreased to 20 nm by reducing the roll-off factor to α = 0, resulting in an SE as high as 3.7 bit/s/Hz even in a non-coherent transmission and at a single-channel bit rate of 10 Tbit/s. The present results will prove fundamental when constructing a future ultrafast TWDM system. By virtue of their high SE and high CD and PMD tolerance, optical Nyquist pulses are expected to provide a simple and direct approach for ultrahigh-speed transmission.
JSPS Grant-in-Aid for Specially Promoted Research (26000009).
We thank Sumitomo Electric Industries, Ltd. and Furukawa Electric Co., Ltd. for providing us with the highly nonlinear fibers used in the pulse compression and the NOLM.
1. G. Raybon, A. Adamiecki, J. Cho, F. Jorge, A. Konczykowska, M. Riet, B. Duval, J.-Y. Dupuy, N. Fontaine, P. J. Winzer, S. Chandrasekhar, and X. Chen, “180-Gbaud All-ETDM single-carrier polarization multiplexed QPSK transmission over 4480 km,” in Optical Fiber Communication Conference2018, paper Th.4.C.3.
2. K. Schuh, F. Buchali, W. Idler, T. A. Eriksson, L. Schmalen, W. Templ, L. Altenhain, U. Dümler, R. Schmid, M. Möller, and K. Engenhardt, “Single carrier 1.2 Tb/s transmission over 300 km with PM-64 QAM at 100 Gbaud,” in Optical Fiber Communication Conference2017, paper Th5.B.5.
3. H. Yamazaki, A. Sano, M. Nagatani, and Y. Miyamoto, “Single-carrier 1-Tb/s PDM-16QAM transmission using high-speed InP MUX-DACs and an integrated OTDM modulator,” Opt. Express 23(10), 12866–12873 (2015). [CrossRef] [PubMed]
4. H. C. H. Mulvad, M. Galili, L. K. Oxenløwe, H. Hu, A. T. Clausen, J. B. Jensen, C. Peucheret, and P. Jeppesen, “Demonstration of 5.1 Tbit/s data capacity on a single-wavelength channel,” Opt. Express 18(2), 1438–1443 (2010). [CrossRef] [PubMed]
5. T. Richter, E. Palushani, C. Schmidt-Langhorst, R. Ludwig, L. Molle, M. Nölle, and C. Schubert, “Transmission of single-channel 16-QAM data signals at terabaud symbol rates,” J. Lightwave Technol. 30(4), 504–511 (2012). [CrossRef]
7. K. Kimura, J. Nitta, M. Yoshida, K. Kasai, T. Hirooka, and M. Nakazawa, “Single-channel 7.68 Tbit/s, 64 QAM coherent Nyquist pulse transmission over 150 km with a spectral efficiency of 9.7 bit/s/Hz,” Opt. Express 26(13), 17418–17428 (2018). [CrossRef] [PubMed]
8. D. Suzuki, K. Harako, T. Hirooka, and M. Nakazawa, “Single-channel 5.12 Tbit/s (1.28 Tbaud) DQPSK transmission over 300 km using non-coherent Nyquist pulses,” Opt. Express 24(26), 29682–29690 (2016). [CrossRef] [PubMed]
9. 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]
10. K. Harako, M. Yoshida, T. Hirooka, and M. Nakazawa, “A 40 GHz, 770 fs regeneratively mode-locked erbium fiber laser operating at 1.6 µm,” IEICE Electron. Express 14(19), 20170829 (2017). [CrossRef]
11. 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 Optical Fiber Communication Conference2005, paper OTuO3.
12. T. Hirooka, K. Harako, P. Guan, and M. Nakazawa, “Second-order PMD-induced crosstalk between polarization-multiplexed signals and its impact on ultrashort optical pulse transmission,” J. Lightwave Technol. 31(5), 809–814 (2013). [CrossRef]