Soliton IEEE Transaction

IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 6, NO.
2, MARCH/APRIL 2000
363
Ultrahigh-Speed Long-Distance TDM and WDM Soliton Transmission Technologies

Masataka Nakazawa, Fellow, IEEE, Hirokazu Kubota, Member, IEEE, Kazunori Suzuki, Eiichi Yamada, and Akio Sahara, Member, IEEE
Invited Paper
AbstractRecent progress on time-division multiplexed (TDM) and wavelength-division multiplexed (WDM) soliton transmission is described, in which dispersion management (DM) plays an important role in increasing the power margin and the dispersion tolerance. The characteristics of the DM soliton are compared with those of return-to-zero (RZ) and nonreturn-to-zero (NRZ) pulses. With a small dispersion swing, the system can still be described as an average soliton with a nonlinear Schrdinger equation (NLSE), whereas with a large dispersion swing, the soliton-like steady-state pulse becomes a chirped Gaussian pulse, in which the master equation is closer to a linear Schrdinger equation (LSE) with a parabolic potential well. An in-line modulation scheme up to 80 Gb/s per channel and its two-channel WDM transmission over 10 000 km are described. A 640-Gb/s (40 Gb/s 16 channels) WDM soliton transmission over 1000 km is also reported with a DM single-mode fiber, without the use of in-line modulation. Finally, dark soliton transmission at 10 Gb/s over 1000 km is described as a different nonlinear pulse application. Index TermsErbium-doped fiber amlifier (EDFA), high speed optical transmission, in-line modulation, optical solitons, time-division multiplexing (TDM), wavelength-division multiplexing (WDM).
I. INTRODUCTION T HAS long been said that a serious drawback to soliton transmission is that a special fiber must be prepared which has an anomalous group velocity dispersion (GVD). However, recently, a new soliton transmission technology has been reported enabling us to use an already installed fiber cable or dispersion-compensated fiber cables. This technology is called the dispersion-managed (DM) soliton technique, by which a soliton can propagate even in fibers with normal GVDs as long as the average GVD is kept in the anomalous region [1][7]. The idea of employing dispersion compensation for soliton transmission was first reported in 1992 to minimize pulse broadening and expand the amplifier spacing [1]. When a DM soliton is used, the sideband instability [8] between the soliton and the dispersive waves has less influence because phase matching is reduced. In addition, the four-wave mixing between WDM channels is also surpressed. Thus, a DM soliton has a much larger power margin than an ordinary soliton under a uniform
Manuscipt received November 24, 1999; revised December 30, 1999. The authors are with NTT Network Innovation Laboratories, Kanagawa-Ken 239-0847, Japan (e-mail: masataka@exa.onlab.ntt.co.jp). Publisher Item Identifier S 1077-260X(00)03447-X.
dispersion [6], [7]. It can be said that the DM soliton is the best way of achieving high-speed, long-distance, single-channel transmission and densely populated wavelength-division-multiplexed (WDM) transmission. The DM soliton does not experience a large nonlinear phase change when the dispersion perturbation is large. The soliton behaves like a linear pulse, such that the pulse broadening caused by positive GVD can be compensated for by negative GVD. However, the DM soliton retains its soliton nature over the average soliton period. The important point is that when the dispersion perturbations are small, the system is fully described by a perturbed nonlinear Schrdinger equation (NLSE), but when the dispersion perturbation is large, the steady-state pulse deviates from being a soliton and is closer to being a Gaussian pulse with chirp [9][16]. The physics of this mechanism is similar to that of FM mode-locking, in other words, stretched pulse mode-locking [17], [18], in which a phase (frequency) modulation is generated by an average self-phase modulation (SPM). Most discussions on DM solitons have been confined to the anomalous GVD region, however, Nijhof et al.has pointed out numerically that a stable, steady-state, nonlinear pulse can exist even in the average normal GVD region [19]. Simultaneously, by using coupled Lagrangian equations for the pulsewidth and corresponding chirp, several authors have pointed out analytically that a stable solution exists even when the average dispersion is in the normal GVD region [20][27]. These nonlinear pulses exist when both the dispersion allocation and the input power are large; however, such pulses are unsuitable for long-distance communication because of strong interaction between them. A rigorous waveform analysis has been made in [28] and [29] by using Hermite Gaussian wavefunctions. In this paper, we describe the superb characteristics of the DM soliton in time-division multiplexed (TDM) and WDM systems. Most of experiments and related analyses in this paper have been undertaken at NTT over the past decade since we invented the 1.48 - m InGaAsP laser diode-pumped erbium-doped fiber amplifier (EDFA) in 1989. We have tried to provide comprehensive explanation of how we were able to establish the DM soliton technique. The readers will easily understand how we could developed this concept. The paper is divided into the following sections.
1077260X/00$10.00 2000 IEEE
364
IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 6, NO. 2, MARCH/APRIL 2000
Fig. 1.
Experimental setup for 80-Gb/s soliton transmission over 500 km.
I II III IV V VI VII
Introduction. Ultrahigh-speed soliton transmission without DM. DM soliton with a small dispersion swing. DM soliton with a large dispersion swing. DM soliton and the stretched-pulse mode-locking. contour map and system evaluation of DM soliton. Experimental comparison of DM soliton, nonreturn-to-zero (NRZ) and return-to-zero (RZ) pulse transmission. VIII Soliton in-line control toward ultrahigh-speed transmission. IX High-capacity, high-speed WDM soliton transmission with DM. X Dark soliton communication. XI Summary. II. ULTRAHIGH-SPEED SOLITON TRANSMISSION WITHOUT DM Before DM was revealed to be a powerful technique for soliton transmission, most researches for solitons had focused on the usage of a dispersion-shifted fiber (DSF) with a slightly anomalous dispersion. Although it was difficult to transmit WDM solitons owing to a solitonsoliton interaction and cross-phase modulation, an ultrahigh-speed single-channel soliton transmission had been already achieved through the use of low-dispersion fibers. In this section, we describe single-pass soliton data transmission experiments of 80 Gb/s over 500 km and 160 Gb/s over 200 km. These experiments were carried out based on the principle of the average soliton, which we called dynamic soliton. Therefore, the transmission fiber has a constant GVD along the transmission distances. A. 80-Gb/s 500-km Soliton Transmission Using Uniform GVD Fiber The experimental setup for an 80-Gb/s 500-km soliton transmission is shown in Fig. 1 [30]. The soliton source was an ac-
tively mode-locked 10-GHz erbium fiber ring laser, capable of emitting a transform-limited 2.73.0-ps soliton pulse at 1.552 m [31]. The pulse was modulated at 10 Gb/s, with a pseudorandom binary sequence (PRBS) using a lithium niobate (LN) intensity modulator. We used a planar lightwave circuit (PLC) as a stable optical multiplexer to obtain an 80-Gb/s pulse train. It consisted of a three-stage MachZehnder interferometer with different arm lengths, which corresponded to time differences of 12.5, 25, and 50 ps. In order to obtain a 10-GHz clock signal easily from the transmitted 80-Gb/s signal, 10-GHz soliton units were superimposed on each other with slightly different soliton amplitudes. This technique is also useful for reducing solitonsoliton interaction [32]. The soliton transmission fibers (STFs) were dispersion-shifted fibers with an average anomalous dispersion of 0.19 ps/km/nm at 1.552 m. The average soliton period was 19.0 km, which meant that the amplifier spacing had to be shortened to as a little as 25 km. The coded pulses were amplified by EDFAs to an average soliton power level of dBm. The average soliton peak power was as high as 31.5 mW. The average fiber loss, including the connector loss for one span, was about 6.0 dB. A narrowband optical filter with a pass band of 3 nm was installed every 50 km to stabilize the soliton train. An 80-Gb/s soliton data signal was demultiplexed to a 10-Gb/s signal using a polarization-insensitive nonlinear optical loop mirror (PI-NOLM) [33]. The demultiplexing circuit is shown in Fig. 2. Part of the transmitted signal was detected with a high-speed InGaAs PIN photodiode, and a 10-GHz clock signal was extracted. We obtained a clock with a high signal-to-noise ratio (SNR) because of the 10-GHz component resulting from our use of an unequal amplitude soliton train. Then, the sinusoidal clock signal drove the 1.533- m distributed feedback laser diode (DFB LD) under a gain-switching condition, and the generated optical pulse was
NAKAZAWA et al.: TDM AND WDM SOLITON TRANSMISSION TECHNOLOGIES
365
Fig. 2.
Optical demultiplexing circuit using a polarization-insensitive nonlinear optical loop mirror (PI-NOLM).
converted to a transform-limited 9-ps pulse with a combination of spectral filtering and linear compression techniques with a 500-m dispersion compensation fiber. The 9-ps pulse train at 10 GHz was amplified and coupled into a part of the loop terminal of the NOLM as a control pulse. The peak power of the control pulse was 198 mW. The NOLM consisted of a 6.0-km long dispersion-shifted polarization-maintained fiber (PANDA) fiber, in which the polarization axis of the fiber was rotated by 90 and spliced to remove the polarization mode dispersion at the middle point of the loop. The timing between the 80-Gb/s signal and the 10-GHz optical clock signal was tuned with an electrical phase shifter and an optical delay. Thus, we obtained a demultiplexed signal at 10 Gb/s. We installed an optical bandpass filter centered at 1.552 m to remove the control pulse at the output and used an EDFA to amplify the demultiplexed signal. We measured the bit error rate (BER) of the demultiplexed signal with an error rate detector. Examples of the performance of the NOLM demultiplexer are shown in Fig. 3, where (a) and (b) show a fixed data pattern at 80 Gb/s before demultiplexing and a demultiplexed 10-Gb/s data pattern, respectively. The waveforms were measured using a streak camera. As the pulsewidth was 2.7 ps, the waveforms did not provide precise information. However, it is clearly seen in (b) that clean demultiplexing has been achieved with a high SNR. When the NOLM gate is off, a ripple is caused by the nonlinear optical bias generated by the counter propagating 80-Gb/s signal when the input signal was in the 1 state. A power penalty is seen in the optical spectrum after transmission; however, the 80-GHz components still remain. This degradation was caused by a slight solitonsoliton interaction. The BER of the demultiplexed signal versus received optical power is shown in Fig. 4, where represents the baseline BER, represents the BER after a 500-km transmission, and shows the BER after a 500-km transmission in a different demultiplexed channel. In this experiment, a power penalty-free soliton transmission was possible over 300 km; however, a power penalty began to accumulate when the transmission distance was ex-
Fig. 3. Demultiplexed signal at 10 Gb/s using a PI-NOLM. (a) 80 Gb/s soliton burst signal at input. (b) Demultiplexed data signal at 10 Gb/s.
tended. The power penalty after a 500-km transmission was 2.53.0 dB and was caused by a slight solitonsoliton interaction. The change of 0.5 dB in the demultiplexed signals is also from the interaction between solitons with unequal amplitudes. The inset photo shows the eye pattern of the demultiplexed 10-Gb/s signal after 500-km transmission, which indicates clear eye opening. A further extension of the maximum transmission distance will be possible by properly choosing the fiber dispersion and pulsewidth. B. 160-Gb/s 225-km Soliton Transmission The transmission speed of 80 Gb/s has been increased to 160 Gb/s using polarization [34] and time-domain multiplexing techniques. The experimental setup for the 160-Gb/s 200-km soliton transmission is similar to that of the 80-Gb/s transmission system shown in Fig. 1 [35]. To reduce the
366
Fig. 4. BER of a 10-Gb/s signal demultiplexed from a transmitted 80-Gb/s signal versus received optical power. baseline; and after 500-km transmission.
The BER of the demultiplexed signal against received optical power is shown in Fig. 5. We obtained a BER of below for both polarized outputs. Although a power penalty occurred of 1.52.0 dB after a 200-km transmission, no indication was given of the existence of an error floor. A power penalty-free soliton transmission was possible over 175 km; however, a power penalty began to accumulate when we extended the transmission distance. The power penalty deviation in the different demultiplexed signals is also from the interaction between solitons with unequal amplitudes. The inset photograph shows a demultiplexed eye pattern after a 200-km transmission, which indicates a clear eye opening. The maximum transmission distance was 225 km with a power penalty of 2.5 dB. Further extension of the maximum transmission distance can be achieved by reducing the average GVD from 0.17 to 0.01-0.02 ps/km/nm, which is also advantageous for reducing the GordonHaus jitter [37]. When a DM technique is used for high-speed soliton systems, one segment of the DM transmission line should be much shorter than the soliton period [38]. This process means that we have to develop a new fiber changing its dispersion in a short period [39]. If this kind of fiber were to become available in the near future, it might prove beneficial for ultrahigh-speed soliton transmission over long distances. Because it is well known that the DM soliton has a large power margin and dispersion tolerance and is useful for reducing solitonsoliton interaction, a DM soliton with rapid DM would be advantageous not only for high-speed TDM, but also for WDM soliton transmission. III. DM SOLITON WITH A SMALL DISPERSION SWING A. Principle In the previous section, we described ultrahigh-speed soliton data transmission without DM. Although soliton transmission is an attractive potential method for realizing high-speed, longdistance communication, serious disadvantages exist such as solitonsoliton interaction, GordonHaus jitter, and change in the GVD along a fiber. In this section, we show that all of the above difficulties can be overcome by adopting DM even in the normal dispersion region, where in principle no bright solitons exist. A DM soliton over the can exist, provided that the average dispersion soliton period is anomalous, even though part of the dispersion in the transmission line is highly normal. It is well known that solitons can withstand dispersion fluctuation within the anomais much longer lous GVD region when the soliton period than the dispersion changeover period [40], [41]. However, no one had yet extended this idea of average solitons to the normal dispersion region [1], [2]. The present idea is that a soliton can remains anomasurvive even in a normal GVD region if lous. Let the distance at which the dispersion value changes from . normal to anomalous or anomalous to normal be represents a dispersion allocation distance shorter than is . Here, we consider a sech-type (hyperbolic secant) bright soliton. That is, we treat here the case of a low-dispersion swing. It is permissible to have a random change in the GVD value between is anomalous. The NLSE normal and anomalous as long as
solitonsoliton interaction between adjacent solitons, the 3-ps pulse from a regeneratively mode-locked erbium-doped fiber laser was compressed to 1.5 ps with a soliton effect [36]. To obtain a 160-Gb/s pulse train, we employed polarization and time-domain multiplexing techniques by using the polarization mode dispersion of a highly birefringent fiber (polarization-maintaining single-mode fiber: PANDA fiber). This process meant that the bit separation was 6.25 ps and the polarization between adjacent solitons was orthogonal to reduce solitonsoliton interaction. The soliton transmission fibers were DSFs with an average dispersion of 0.17 ps/km/nm at 1.552 m. The average soliton period was 5.5 km, and the amplifier spacing was 25 km. This condition violates the principle of the average soliton because the perturbation occurs in a shorter distance than does the soliton period. Stable transmission over longer distances would be possible if the amplifier spacing were reduced to less than 10 km. However, this reduction would not be practical because an EDFA can be installed with a much longer span because of its low-noise and high-gain characteristics. The soliton peak power was as high as 109 mW. average We installed a narrowband filter with a passband of 5 nm every 50 km to stabilize the soliton energy. In the optical demultiplexing, the 160-Gb/s soliton data signal was first demultiplexed to 80 Gb/s using a polarization beam splitter, and then further demultiplexed to a 10-Gb/s signal using a PI-NOLM as shown in Fig. 2.
367
(4) Here, we assume that one repeater span consists of two fibers. with a length of , and the other One fiber has a GVD of with a length of . has a GVD of is given as By definition, (5) and (4) is rewritten as
(6) In order for to behave as a soliton, should be time indeshould be pendent, and therefore, the coefficient of zero. This requirement gives us the following relationship beand the fiber loss : tween the initial amplitude (7) The average soliton period,
Fig. 5. BER of a 10-Gb/s signal demultiplexed from a transmitted 160-Gb/s signal versus received optical power. baseline; after 200-km transmission; after 225-km transmission.
, is expressed as (8)
with a perturbation, , and a dispersion, , which is dif, is given by ferent from the average dispersion, (1) Assuming that can be described in the form (2a) and satisfy the following equations and that when an average soliton experiences a fiber loss, as a perturbation: (2b) (2c) one obtains (3a) (3b) Thus, the total nonlinear phase change for a repeater spacing is given by
can be extended much where is the input pulsewidth. , even though we use highly normal or highly further than anomalous GVD fibers that usually give short soliton periods can be extended in each section. By choosing a small to a distance comparable with that in a conventional intensity modulated direct detection (IM-DD) system. For a pulsewidth of , the soliton peak power in the present system is given by (9) The DM soliton amplitude at the input is the same as that of an average soliton. A significant difference between the present soliton and a conventional average soliton is that it can propagate even in a normal GVD region. A stable soliton exists when the product of the fiber dispersion and its length is much smaller than
(10) is the dispersion perturbation and is the norHere, was coined parameter, which exmalized distance. This
368
presses a strength of the dispersion map [19], [27]. As is much smaller than unity, is an important requirement for transmitting a DM soliton with a sech waveform. The DM soliton does not experience a large nonlinear phase change when the above requirement is satisfied, where the soliton behaves like a linear pulse such that the pulse broadening caused by the normal GVD can be compensated for by the anomalous GVD. However, the DM soliton has the na. ture of a pure soliton over Fig. 6 shows how DM solitons can propagate over long is kept to 90 km and was 30 distances [2]. Here, km in the normal GVD region and 60 km in the anomalous region. The transmission speed was 10 Gb/s, the input DM soliton pulsewidth was 20 ps, and the transmission distance pseudorandom pattern is used as the was 1170 km. A input soliton train, and the results show eye patterns at the output. In Fig. 6(a), the soliton propagates from a 30-km fiber ps/km/nm to a 60-km fiber with a GVD of with a GVD of ps/km/nm. This condition gives a of 0.2 ps/km/nm of 792 km, which is much longer than . As and a seen in Fig. 6(a), the soliton was able to propagate stably over 1170 km easily without distortion. is made normal. To show that this pulse was a soliton, was Fig. 6(b) shows a nonsoliton transmission in which ps/km/nm. Here, we used a 30-km fiber with a set at ps/km/nm GVD and a 60-km fiber with a ps/km/nm GVD. As can be clearly seen, the pulse was distorted and eventually became rectangular. These results indicate that an average is anomalous. soliton can exist when Fig. 7 shows how the soliton pulsewidth changes as it propagates through a DM fiber consisting of a 30-km fiber ps/km/nm GVD and a 60-km fiber with a with a ps/km/nm GVD. is 0.2 ps/km/nm, and is 793 km. The pulsewidth of the present soliton (thick solid line) deviates slightly from that of the uniform average soliton (dotted line) because the dispersion perturbation is large, but it is entirely different from a linear pulse, as shown by the dashed line. The thin solid line represents the situation when a smaller dispersion ps/km/nm GVD and perturbation (a 30-km fiber with a a 60-km fiber with a 0.5 ps/km/nm GVD) is applied to the system, in which more stable pulse propagation is achieved. The dotted line shows the condition when a uniform GVD of ps/km/nm is used. These results prove that a new kind of soliton exists in a DM fiber system, and this enables us to use different kinds of fiber. This technique makes it possible not only to extend the repeater spacing [1], but also to reduce GordonHaus jitter as the average GVD can be set at a small value [3]. It is also possible to control the new solitons using filters, modulators, and optical gain. Here, we analyze how much perturbation can be allowed for is much larger than , the stable soliton propagation. As average soliton behaves just like a dispersive wave in each recaused by gion. Thus, the pulse broadening is
Fig. 6. Soliton transmission through DM fiber. (a) A DM fiber with a D of : ps/km/nm consisting of a 30-km, ps/km/nm section and a 60-km, : ps/km/nm section. (b) From -ps/km/nm fiber to : -ps/km/nm fiber with a D of : ps/km/nm.
+0 2 +1 3
00 2
02
02
+0 7
Fig. 7. Change in pulsewidth of DM soliton against propagation distance. Dashed line: linear pulse; dotted line: uniform average soliton; thick solid line: DM soliton with larger dispertion perturbation; thin solid line: result with smaller dispersion perturbation.
An important requirement for the present soliton is that this broadening should be much less than the original pulsewidth. Thus, a requirement of
(12) is obtained. This result is reasonable because if the broadening is not negligible in relation to the original pulsewidth, the perturbation to the soliton becomes large and eventually the soliton pulse becomes a mere dispersive wave. A soliton is stable when the product of dispersion and fiber length is smaller . Here, is the normalized distance than . We will that has a relationship of of much larger than unity in describe the DM soliton with Section IV, in which a larger peak power is required to maintain such solitons.
(11)
369
B. Soliton Field Experiments Using Commercially Installed Optical Fiber Cables with a Random GVD Variation The dispersion allocation technique described in the previous section made it possible to undertake a soliton communication field trial easily using the conventional fiber cable already installed for commercial 2.4-Gb/s IM-DD systems. In this section, we describe the first soliton transmission field test, in which we used part of the Tokyo metropolitan optical loop network with a radius of 100 km [42]. 1) 10-Gb/s 2500-km and 20-Gb/s 2000-km Transmission: The location of the field test is shown in Fig. 8(a). We used a total of four repeater terminals, which were located at Mito, Utsunomiya, Sano, and Maebashi. The distances between the Mito and Utsunomiya (MU) terminals, Utsunomiya and Sano (US) terminals, and Sano and Maebashi (SM) terminals were 70, 50, and 50 km, respectively. The transmitter and receiver were placed at the MU terminal, and the total transmission distance was extended to 2500 km by repeatedly cascading these terminal spans as shown in Fig. 8(b). The average fiber losses of the MU, US, and SM sections were 18. 5, 14.3, and 14. 3 dB, respectively. Two kinds of optical soliton source were used for the field test. One source was a regeneratively mode-locked fiber laser [31] operating at 10 GHz with a pulsewidth of 13 ps and a spectral width of 0.27 nm. The other source was a gain-switched multiple quantum-well (MQW) DFB LD operating at 1.552 m. A 13-ps transform-limited pulse was generated with gain switching and spectral windowing techniques [43]. The pulse was modulated at 10 Gb/s with a PRBS using an LN modulator. Each EDFA consisted of a 1.48- m InGaAsP LD for pumping, a WDM coupler, an isolator, an erbium fiber, and a 3-nm optical filter. The optical cable in the network we employed is normally used for a 2.4-Gb/s commercial system. We used 36 of the 200 cores (fibers). In each fiber cable, one piece was typically 2 km ps/km/nm. As in length with a different GVD of less than shown in Section V, a stable soliton can exist when the product of the fiber dispersion and its length for each 2-km fiber piece [2], [4]. A DM soliton has the is much smaller than . nature of a pure soliton over In each section, we measured the GVD of 36 fibers with the roundtrip phase rotation method we developed [36]. With this method, the signal wave is reflected at the far end of the fiber by a mirror with a high reflection or optical circulator, and it is amplified by an EDFA installed at the far end. This method enables the GVD of an installed fiber to be measured easily and precisely over 100 km. When the soliton wavelength is set at 1.552 m, more than 50% of the fibers can be used for soliton transmission without employing dispersion allocation. Here, we chose 16 fibers in each section and constructed a 2500-km soliton transmission line, in which the GVD varied between 0.05 and 0.4 ps/km/nm in the anomalous region without employing dispersion allocation. The average GVD was approximately 0.25 ps/km/nm, and the average soliton period was 364 km. It is important to note that 100% fibers can be applicable to DM soliton transmission when a proper dispersion compensation fiber is attached to keep the GVD to be anomalous.
Fig. 8. Experimental field transmission of solitons at 10 Gb/s over 2500 km using part of the Tokyo metropolitan optical loop network with a radius of 100 km: (a) location of field test and (b) soliton transmission route.
The input and output waveforms (2500 km) of a pseudorandom soliton pulse and their spectra are shown in Fig. 9(a) and (b) are the input waveform and its spectrum and (c) and (d) are the output waveform and its spectrum. The soliton source was a fiber laser. It is surprising to note that a 13-ps input pulse remained unchanged even after a transmission of 2500 km through installed dispersion shifted fiber cables. A degradation occurred in the SNR after 2500 km, but the eye pattern was clearly open, as shown in the inset photo in Fig. 10. Fig. 10 shows the BER characteristics measured before (back-to-back) and after a 2500-km transmission at 10 Gb/s. Error-free operation was successfully achieved at a received dBm. The power penalties after transmissions of power of , 2040 and 2520 km were 3.1 and 5.8 dB at a BER of respectively. This penalty may have been caused by dispersion-related dispersive wave perturbation acting on the solitons. We also examined a gain-switched LD as a soliton source. No significant difference existed between the waveforms and their spectra with the two soliton sources. The power penalty after the 2040-km transmission was 3.0 dB and no change occurred at all in the BER characteristics compared with that of the fiber laser source. A further extension of the error-free transmission distance will be possible through proper dispersion allocation, which can be realized by adding dispersion compensation fibers [4].
370
Fig. 9. Input and output waveforms and corresponding spectra: (a) input PR soliton waveform, (b) spectrum, (c) output PR soliton waveform, and (d) spectrum.
We also succeeded in a 20-Gb/s soliton transmission using the same optical transmission line shown in Fig. 8 [5]. Fig. 11 shows the BER characteristics measured before (back-to-back) and after a 2040-km transmission at 20 Gb/s. Error-free operation at 20 Gb/s was successfully achieved at a received power dBm. The power penalty after a 2000-km transmission of . The lower average dispersion was 3.5 dB at a BER of results in the higher speeds and longer transmission distances. 2) 40-Gb/s 1000-km Soliton Field Experiment with Dispersion Management: In this section, we describe a 40-Gb/s soliton transmission field experiment over 1000 km, which we undertook in the optical network described in the previous section. Here, we employed a dispersion-compensation technique to keep the average dispersion at a small constant value, and this resulted in an error-free 5-ps soliton data transmission over 1000 km. One section of fiber cable was typically 2 km long, and the ps/km/nm. We chose fibers in each section had GVD of six fibers in each span and constructed a 1000-km soliton transmission line. The average fiber dispersion for one repeater section at 1552 nm varied between 0.06 and 0.23 ps/km/nm. We adjusted the average GVD to 0.05 ps/km/nm by inserting a dispersion-compensation fiber (DCF) before each EDFA repeater. ps/km/nm. The DCF had a GVD of approximately The optical source was a regeneratively mode-locked fiber laser operating at 10 GHz with a pulsewidth of 5 ps and a spectral width of 0.5 nm. The pulse was modulated at 10 Gb/s with PRBS using an LN intensity modulator. A PLC was a
used to optically multiplex the 10-Gb/s signal into a 40-Gb/s data train. The average fiber-coupled signal power for the MU, US, and SM sections were , and dBm, respectively. The transmitted 40-Gb/s soliton pulse was demultiplexed directly into a 10-Gb/s signal by using an electroabsorption modulator with the recovered 10-GHz clock signal obtained from part of the transmitted 40-Gb/s signal. The gate width for the optical demultiplexing was 20 ps, and the extinction ratio of the gate switch was more than 20 dB. We detected the demultiplexed signals with an InGaAs optical receiver and then measured the BER for each demultiplexed channel. The input and output waveforms (1020 km) of a pseudorandom soliton pulse and their spectra are shown in Fig. 12. Fig. 12(a) and (b) show the input waveform and its spectrum and Fig. 12(c) and (d) show the output waveform and its spectrum, respectively. Pulse broadening of about 1 ps occurred during transmission. However, we observed a timing shift of approximately 2 ps from proper pulse position. As the amount of timing shift depended on the polarization state of the soliton signal, this shift was mainly caused by residual polarization mode dispersion of the fiber cable. We observed SNR degradation in the pulse waveform after a transmission of 1020 km. Because the optical spectra after the transmission had an SNR of more than 30 dB, the DC component in the waveform was mainly from nonsoliton components. In Fig. 12(d), the spectral component corresponding to a 40-GHz carrier was reduced after the transmission, which may be caused by the timing shift of the transmitted soliton pulse.
371
Fig. 10. BER at 10 Gb/s over 2500 km versus received optical power. baseline; 4 after 2040-km transmission; after 2520-km transmission.
Fig. 11. BER characteristics at 20 Gb/s over 1020 km. baseline; 4 and after 1020-km transmission.
Fig. 13 shows the BER characteristics at 40 Gb/s measured before (back-to-back) and after a 1020-km transmission. The open and closed circles shows the BER after 1020-km transmission in channels 1 and 2, respectively. The open and closed triangles shows the BER in channels 3 and 4, respectively. Error-free operation was successfully achieved at a received dBm. The power penalty after the transmission power of . This penalty may was 1.7 to 2.4 dB at a BER of have been caused by the residual polarization mode dispersion (PMD) and accumulated dispersive waves, as shown in Fig. 12(c). This 40-Gb/s soliton field experiment was important because it verified that data can be sent at such a high speed over 1000 km through an already installed DSF that was not designed for such high-speed data transmission. IV. DM SOLITON WITH A LARGE DISPERSION SWING A. Principle Many numerical and analytical reports have been published with respect to DM solitons [1][29]. When a DM soliton is used, we employ fibers with a normal GVD and an anomalous GVD alternately for a simplest case, so that the sideband instability between the soliton and the dispersive waves has less influence because phase matching, which will generate the FWM, is unlikely to occur. In other words, the DM solitons are useful not only for compensating for the GVD, but also for eliminating other nonlinear effects developing as a result of the phase
matching caused by uniform low dispersion over long distances. When the dispersion perturbations are small, the DM system is fully described by a perturbed NLSE, as we described in the previous section. However, when the dispersion perturbation is large, the steady-state pulse deviates from being a sech soliton and is closer to being a Gaussian pulse with a chirp. Here, we describe the case of a large dispersion swing. Most discussions on DM solitons have been confined to the anomalous GVD region; however, it has been numerically pointed out that a stable nonlinear pulse exists even in the average normal GVD region [19]. Simultaneously, by using coupled Lagrangian equations concerning the pulsewidth and corresponding chirping, several authors pointed out analytically that a stable solution exists even when the average dispersion is in the normal GVD region [20][26]. Such a nonlinear pulse exists when both the dispersion allocation and the input power are large. To investigate the steady-state pulse characteristics, we here use the variational method derived by Anderson et al. [46], [9][26]. Because of a large linear chirp in the transmission line, which is of course modified by the Kerr nonlinearity, we assume a Gaussian ansatz as the steady-state pulse (13) , and indicate the pulse amplitude, pulsewidth, where chirp and phase, respectively. By putting (13) into an original NLSE without fiber loss (1), the variational method yields the
372
Fig. 12. Input and output waveforms and corresponding spectra: (a) input pseudorandom soliton waveform, (b) input soliton spectrum, (c) output pseudorandom soliton waveform, and (d) output soliton spectrum.
following coupled equations between the pulsewidth and the chirp:
(14a) (14b) , and where is the normalized dispersion, (energy is constant). We neglect here a fiber loss. In Fig. 14, we describe the well-known relationship between , and the normalized pulse the normalized average GVD, energy by using the normalized dispersion allocation given by . Therefore, an actual normalized GVD at the higher GVD and that at the lower region is given by . The length for GVD region is given by . each GVD region is the same, so that values of 2, 26.5, and 50 are used in the figure. of 2 of 26.5 means that means dispersion management is small. of the lower GVD well gets into the normal GVD region. of 50 indicates that the system is largely DM. For a small
2, the pulse energy increases linearly with the average dispersion, which is a characteristic of a sech-shaped soliton pulse (low-dispersion swing). This average soliton has already been described in the previous section. When the DM becomes large with a of 26.5, the soliton energy is increased to maintain it as a steady-state pulse. We call this energy increase an energy enhancement factor. However, in this dispersion allocation map, no steady-state pulse still exists in the average normal dispersion region. When we further increase the dispersion swing to a of 50, a nonlinear pulse starts to have a steady-state solution even in the normal GVD region. Such behavior starts to occur at of around 30. This process can be understood as follows. a In the first segment of the fiber (anomalous GVD), soliton compression occurs because of the large pulse peak power, resulting in a slight spectral broadening. When such a pulse is coupled into the second segment of the fiber (normal GVD), the pulse , is broadened further compared with a linear case (large but small coupled power). With this pulse change, a steady-state minimum pulse in the anomalous GVD section is narrower than one in the normal GVD section. In other words, the dymanic average GVD defined by GVD/(pulsewidth) integrated over the anomalous GVD region becomes bigger than that in the normal
373
Fig. 14. Relationship between normalized energy " and path-average dispersion D .
Fig. 13.
BER at 40 Gb/s over 1020 km versus received optical power.
GVD region, even when the absolute GVD in the anomalous GVD region is smaller than that in the normal GVD region. That is, the dynamic average dispersion is still in the anomalous GVD region, which is the first principle of the existence of a steady-state pulse in the normal GVD region. However, it should be noted that the pulse is inevitably and greatly broadened in one period of the DM fiber. When we try to use such a steady-state nonlinear pulse train in the normal GVD region for a high-speed optical transmission, an unusually large nonlinear interaction occurs. This process makes it difficult to send a data signal over long distances, which occurs because of the greatly broadened nonlinear pulses in the normal GVD region, where pulse interaction is significant because the coupled peak power of the pulse still generates a nonlinear phase change. In Fig. 15, we show typical pulsepair interactions in three cases. In each case, one span consists of a 50-km long fiber with an anomalous GVD of 16 ps/km/nm and a 10-km long DCF with a normal GVD. The normal GVD was changed to obtain a different GVD with a different sign. The pulsewidth at the input and the separation were 20 and 100 ps, respectively. In Fig. 15(a)(c), the average GVD was set at 0.1, 0.0, and 0.2 ps/km/nm, respectively. The corresponding coupled powers are 2, 0, and 14 dBm (corresponding peak powers: 12, 10, and 24 dBm), respectively. That is, Fig. 15(a) is a pure DM soliton in the anomalous GVD region. Fig. 15(b) is a zero GVD pulse. Fig. 15(c) corresponds to soliton-like pulse propagation in the normal GVD region. Therefore, the coupled power in Fig. 15(c) is much larger than those in Fig. 15(a) and (b). The pulse pair maintains its separation in Fig. 15(a), although inter-
ference occurs between the pulses. In Fig. 15(b), a strong interaction occurs, but the pulse separation does not change greatly. However, it should be noted that a large pulse separation occurs in Fig. 15(c). This result indicates that such a soliton-like pulse in the normal GVD region has a strong interaction force. This force occurs because of its large transmitting power. Even when the pulse is broadened because of the large GVD, a nonlinear phase change still occurs. In Fig. 15(c), although the transmitting pulsewidth is much shorter than its pulse separation (the pulse separation when the interaction occurs is more than 10 times broader than the pulsewidth), a nonlinear interaction still occurs because of the large pulse broadening accompanied by a sufficient nonlinear phase rotation in one span. When pulse overlapping occurs, it causes a mutual pushing effect, which leads to a large residual pulse separation, as shown in Fig. 15(c). These results suggest that we must be careful to use a nonlinear steady-state pulse in the normal GVD region on the basis of larger pulse interactions. The actual transmission quality including the behavior of these pulses can be evaluated numerically by using a mapping technique we developed [47]. This technique will be described in Section VI. V. DM SOLITON AND THE STRETCHED-PULSE MODE-LOCKING We describe here the relationship between the DM soliton and the stretched-pulse mode-locked fiber laser [13], [14], [17], [18]. The pulse propagation in a soliton laser corresponds to an ideal or average soliton transmission. It has an anomalous GVD, and the pulse maintains a constant width and remains transform-limited over a single pass. By contrast, the pulse propagation in a stretched-pulse mode-locked laser corresponds to a DM soliton transmission, in which the net GVD is close to zero, and the pulse undergoes large dispersive changes in width and peak power in a single pass. The master equation for the pulse propagation in a soliton laser, which corresponds to a uniform soliton transmission, is written in the following form with a perturbed nonlinear Schrdinger equation:
(15)
374
Fig. 15. Soliton collision characteristics: (a) average GVD 0.1 ps/km/nm, coupled peak power power 10 dBm; and (c) average GVD 0.2 ps/km/nm, coupled peak power 24 dBm.
=0
= 12 dBm; (b) average GVD = 0 ps/km/nm, coupled peak
It is well known that the solution of (15) is a chirped sech pulse, as shown by (16) On the other hand, the master equation for a stretched pulse mode-locked laser or a DM soliton is described with a linear Schrdinger equation (LSE) as
VI.
CONTOUR MAP AND EVALUATION OF DM SOLITON TRANSMISSION SYSTEM
(17) In the stretched-pulse laser, the pulsewidth is greatly modified and the average pulsewidth is broader than the minimum width. Therefore, the field experiences average nonlinearity, and we can approximate the nonlinear potential with a parabola. Thus, we obtain the above equation. The solution of (17) is also well known and given by (18) is the complex and the DM soliton has a Gaussian Here, rather than a sech pulse shape. With the DM soliton, is zero in (17). Haus and Chen recently showed that the essential physical mechanism of the DM soliton can be explained by the interacand HermiteGauss components of tion of the the pulses ([28] and [29] and references therein).
The value is the SNR of the received data eye and is directly related to the BER of the transmission system [46]. A value is calculated from fluctuations in the transmitted waveform (eye pattern), as shown in Fig. 16. If we assume the fluctuation has a Gaussian distribution, the BER can be directly calcorresponds to culated from the value. For example, . Although the value is a measure of the a BER of SNR, it can also include a degradation in the transmission characteristics because of a nonlinear interaction of the pulse when a PRBS is used as the input data stream. Therefore, we can evaluate the transmission characteristics of a complicated transmission system that has a nonlinear interaction and noises simply from the value. An unnormalized NLSE with fiber loss and dispersion change as a function of distance is given by
(19) is the normalized local GVD value and is the where corresponds to the average GVD. The normalized fiber loss. is assumed last term indicates the self Raman effect, where to be 5.9 fs. The amplified spontaneous emission (ASE) power PASE, generated by the EDFA, is given by (20) where is the photon energy; is the bandwidth; is the gain; is an inversion parameter.
375
The ASE is assumed to be white noise and added to the signal after each amplification process. An optical filter is also applied to the transmitted waveform after the ASE has been added. The transmitted signal is detected by a quadrature detector. A baseband electrical filter of 0.65 bandwidth is applied to the detected signal to reduce the noise, and then an eye pattern is drawn to measure the value. Because this method is applicable to any waveform and any dispersion configuration, we can compare the transmission characteristics of various transmission systems from the value. Using this value, we compare the characteristics of the DM soliton with those of conventional NRZ and RZ pulse transmissions at zero GVD. Fig. 17 shows a new method for evaluating the performance of optical transmission systems at 20 Gb/s. This method is called factor contour mapping and can immediately evaluate the power margin and dispersion tolerance of a system [47]. With this map, we can compare the transmission performance of different optical systems. Fig. 17(a) and (b) shows contour maps of the transmitted peak optical power (EDFA output power) and the GVD of a transmission fiber. The transmission distance is set at 2560 km. This mapping enables us to evaluate the power margin from the vertical axis and the dispersion tolerance from the horizontal axis. Fig. 17(a) and (b) corresponds to the soliton and the NRZ transmission, respectively. In both cases, the GVD is uniform. The contours of the map in Fig. 17(b) are entirely different contour is much smaller from those in (a), in which the than that in (a). It is interesting to note that the map is not symmetric in relation to zero GVD, which would be expected from a linear transmission property, but the results suggest that better transmission quality is obtained in the anomalous GVD region. This result means that some type of soliton effect improves the transmission quality even for NRZ pulses. These results imply that a soliton transmission with uniform GVD is of better quality than an NRZ transmission at exactly zero GVD. map Fig. 17(c) and (d) shows how the contours of the can be improved through the use of dispersion allocation. The transmission distance is set at 5120 km in (c) and at 2560 km in (d). As seen in (c), the DM soliton can drastically extend the transmission distance to over 5120 km, which is much further than that of a soliton with uniform GVD. The can be obtained at a much higher transmission maximum power corresponding to a power enhancement factor [6], [7]. These improvements are realized because phase matching is unlikely to occur. The transmission distance of an NRZ pulse using dispersion allocation is also further than that in (b), ), which but it is limited to 3200 km (dotted line with is much shorter than that of the DM soliton. As clearly seen in (c), DM soliton transmission has a much larger power margin and dispersion tolerance than with DM RZ and NRZ pulse transmissions at zero GVD. The contours in (c) also show that the dispersion tolerance is much larger than that of the NRZ system. These results suggest that DM soliton transmission is a powerful optical transmission technique. DM solitons are useful not only to compensate for the GVD, but also to eliminate other nonlinear effects occurring as a result of the phase mismatching caused by the large GVD in each fiber segment.
Fig. 16.
Definition of the
Q value.
Fig. 18 compares the performance of three optical transmission systems (DM soliton, NRZ and RZ transmission zero GVD) numerically. Two-step dispersion allocation is used in these simulations. Fig. 18(a) and (b) show results for 10- and , is the 20-Gb/s systems, respectively. The horizontal axis, amount of dispersion allocation from the average GVD in the first fiber segment with anomalous GVD and the vertical axis is the transmission distance. The maximum transmission distance factor becomes seven. is defined as the point at which the For the 10-Gb/s evaluation, we used a 60-km span consisting of a 50-km long anomalous GVD fiber and a 10-km dispersion allocation fiber. The amplifier noise figure was 8 dB, PRBS , and the pulse width was 15 ps. The full width at was half maximum (FWHM) of the filter width was 3 nm for both the DM soliton and NRZ at 10 Gb/s, and the band width of the baseband filter was 6.5 GHz. For the 20-Gb/s evaluation, we employed different parameters. Here, one span was 80 km. The first fiber had an anomalous GVD and was 30 km long, and the succeeding fiber had a normal GVD and was 50 km long. The pulsewidth was 10 ps, and the baseband filter width was 13 GHz. The FWHM of the installed optical filter was 2 nm for the DM soliton and 10 nm for the NRZ pulse. Our intention was to derive a fundamental principle for two-step dispersion allocation even under different conditions. The various transmitted powers were chosen to obtain the maximum transmission distance for each dispersion allocation. As it is known that modulational instability degrades the quality of an NRZ transmission [48], the dispersion allocation was reversed; that is, the pulse train was transmitted from a normal to an anomalous GVD fiber. It is clear from Fig. 18 that the DM soliton has the longest transmission distance in both 10- and 20-Gb/s systems. It is interesting to note that the transmission distance is the most sensitive quantity of the dispersion allocation in the soliton system. For example, the maximum distance is obtained at a dispersion allocation of approximately 46 ps/km/nm for 10 Gb/s and 23 ps/km/nm for 20 Gb/s, which means that an optimum allocation exists at which the average Kerr effect can be stably balanced with the average GVD. Even for RZ transmission at zero GVD, the maximum distance depends on the dispersion allocation, because the soliton nature can be maintained in the anomalous GVD fiber, although the overall soliton effect is weak because the average GVD over one span is zero. For NRZ transmission, the maximum transmission distance is almost independent of dispersion allocation, which indicates that NRZ is at its most stable in the linear transmission region under relatively low transmitted power. From Fig. 18(a) and (b), the typical transmitted powers needed to obtain the maximum transmission dis-
376
Fig. 17. contour maps for various optical transmission systems: (a) soliton with uniform GVD, (b) NRZ pulsewith uniform GVD, (c) DM soliton, and (d) DM NRZ pulse.
tance for DM soliton, RZ pulse, and NRZ pulse transmission at , and dBm, respectively and those 10 Gb/s were , and dBm, respectively. at 20 Gb/s were These results indicate that the DM soliton has the largest transmitted power and suppresses FWM with the aid of the DM. Finally, we compare the reduction in the transmission distance for DM soliton and NRZ transmission when the average dispersion is changed. If the change in the transmission distance is sensitive with respect to average dispersion, it may be difficult to realize a commercial system, hence, the importance of this present analysis. The result is shown in Fig. 19 for DM soliton using 1.3- m single-mode fibers (SMFs). The conditions are the same as those of Fig. 18. The launched power was optimized for each dispersion. The arrows in the figure indicate the transmission distance at a dispersion tolerance of 0.4 ps/km/nm. The distance is about 5000 km for a DM soliton pulse and 2000 km for an NRZ signal. Here again, the transmission distance of the soliton signal is about 2.5 times greater than that of the NRZ signal against the change in the average GVD. These results indicate that a high-quality transmission with a larger margin is possible with the DM soliton. VII. EXPERIMENTAL COMPARISON OF DM SOLITON, NRZ, AND RZ PULSE TRANSMISSION We confirmed the results shown in the previous section with transmission experiments using a fiber loop, as shown in Fig. 20.
The experiments comprised a 10-Gb/s data transmission using SMF and DCF [49]. The amplifier spacing was 59 km, where a 50-km SMF was succeeded by a 9-km DCF, and the loop length was 472 km. The average GVD was 0.1 ps/km/nm. This condition corresponds to the filled squares on the dotted line in Fig. 18(a). The pulsewidth was 16 ps, and the corresponding spectral width was 0.21 nm. The time-bandwidth product was 0.42. The signal wavelength was 1.553 m, and the data length PRBS. The fiber loop consisted of eight amplifier was spans, with eight 0.98- mpumped EDFAs and one 1.48- m pumped EDFA. In Fig. 18(a), the dispersion allocation with the 1.3- m fiber corresponds to as much as 16 ps/km/nm; nevertheless, this process still contributes to extending the transmission distance, which is why we used conventional 1.3- m fiber. This technique is applicable to DSFs in which much better results can be obtained as expected from Fig. 18 because the dispersion allocation is much smaller than that for 1.3- m SMF. Fig. 21 shows BER results for solitons and RZ and NRZ pulses at zero GVD. The transmission distances for the NRZ, RZ, and DM soliton systems, which guaranteed a BER of , were 3800, 4800, and 6000 km, respectively. As clearly seen, the longest transmission distance was achieved with the DM soliton. This result agrees well with the numerical analysis [see at 16 ps/km/nm in Fig. 18(a)]. In Fig. 18(a), the DM soliton has the longest transmission distance of 8000 km, next is the RZ pulse at zero GVD (5000 km), and then the NRZ pulse in all cases. These (2500 km). The BER was set at
377
Fig. 19.
Dispersion tolerance of DM soliton and NRZ systems.
Fig. 18. The maximum transmission distance versus two-step dispersion allocation. is the amount of dispersion allocation from the average GVD in the fiber segment: (a) 10-Gb/s system and (b) 20-Gb/s system.
1D
results agree well with the numerical results. The difference between the experiment and the numerical analysis may be from the polarization dependence of the gain and loss in the system and small parameter differences. It is interesting to note that the input pulses were immediately broadened to as much as 180 ps by the dispersion; nevertheless, the results indicate that an accumulation of a small amount of Kerr nonlinearity exists which balances to the average dispersion giving a value of as small as 0.1 ps/km/nm. We were able to extend the maximum transmission distance for 1.3- m DM soliton transmission even further by incorporating a slow-speed polarization scrambler to eliminate the polarization hole burning effect in the EDFAs. BER curves before and after the 10 000 km transmission are shown in Fig. 22.
A BER below was achieved with a received optical dBm. The power penalty at that point was 5.2 dB. power of The inset photograph shows the eye diagram after a 10 000-km transmission, which indicates a clear eye opening. We investigated the BER as a function of the transmission distance and at 10 600 km. This result agrees with achieved a BER of our numerical simulation, which gives a maximum transmission distance of more than 11 000 km (see Fig. 19). Even when the amplifier spacing was extended to 59 km (50-km SMF and 9-km at a transmission DCF), we achieved a BER below distance of 9200 km. Excellent experiments at 1020 Gb/s over 20 000 km have also been successfully reported by Carter et al. and Doran et al., in which they showed that DM solitons using SMF is the best way of upgrading SMF links [50][54]. An alternative dispersion compensation technique to DCF is a chirped fiber Bragg grating (CFBG) [55], [56]. CFBGs are more compact and less expensive than DCFs, and they do not suffer from nonlinear effects in the fiber. Recently, group delay ripples appearing on the dispersion curve of a CFBG have been reduced, and therefore, we can further improve the transmission characteristics of DM solitons using SMFs and CFBGs. We have succeeded for the first time in sending a 10-Gb/s soliton signal over 2900 km using 1.3- m SMFCFBG dispersion compensation [57]. The chirped fiber gratings we used were approximately 100 mm long, with a 0.8-nm bandwidth, and 850ps/nm total dispersion. VIII. SOLITON IN-LINE CONTROL TOWARD ULTRAHIGH-SPEED TRANSMISSION A. Principles in Soliton Transmission Control In a linear system, the GVD broadens the pulsewidth and fiber loss reduces the signal intensity. If we use optical solitons as the signals, the dispersion problem is overcome and the
378
Fig. 20. Experimental setup for comparing systems using 1.3-m SMFs and DCFs.
Fig. 21. BER versus transmission distance for DM soliton, RZ, and NRZ pulses. : DM soliton; : RZ pulse at zero GVD; 4: NRZ pulse at zero GVD.
loss is compensated for by EDFAs. However, the use of solitons causes new problems, such as solitonsoliton interaction, GordonHaus jitter [38], and the accumulation of ASE noise. GordonHaus jitter has a cubic dependence on the propagation distance and is a major factor in limiting soliton transmission distances [38]. The variance in the jitter is given by (21) is where is amplifier gain expressed by is amplifier spacing, is Plancks connormalized fiber loss, is the nonlinear index, is GVD, is transmission stant, is the effective distance, is the FWHM pulsewidth, and area. For example, for a pulsewidth of 30 ps, a repeater spacing ps/km/nm, and of 50 km, a loss of 0.25 dB/km, a GVD of a spot size of 4 m, the GordonHaus jitter is as large as 19 ps
Fig. 22. BER curves before and after 10 000-km transmission. : 0 km; 10 000 km.
after a propagation of 10 000 km. The limit of the bit rate and , is given by distance product, which provides a BER of (22) is the detector where is the bit rate of the system and Gb/s and window width. For example, when ps, the maximum transmission distance is only 7800 km. To ensure a smaller BER, the total transmission distance should be shortened.
379
We proposed a way to overcome the GordonHaus limit, ASE accumulation, and solitonsoliton interaction [58][61]. This technique is called soliton control or in-line modulation and was used to achieve a 1-millionkm soliton transmission [58]. Soliton control is similar to active or passive mode-locking techniques. In the active and passive mode-locking of lasers, synchronous modulation is used for pulse shaping and retiming, and a saturable absorber is used for noise reduction and pulse shaping. The pulse shaping and retiming techniques are useful for soliton control. Synchronous modulation (soliton control in the time domain) enables us to retime the position of the soliton pulse, which experiences jitter as a result of ASE noise. It is also possible to remove the interaction forces inevitably occurring between closely adjacent solitons. We proved theoretically that soliton data transmission over unlimited distances is possible with this technique because periodic synchronous modulation can reduce the ASE noise to a low level [60]. We also installed a bandpass filter with a 0.30.4-nm bandwidth in the loop as a soliton control in the frequency domain to stabilize the soliton energy [59], [60]. Fig. 23 shows the principle of noise reduction through the use of synchronous modulation [59]. When modulation is applied to a soliton, it can be reshaped and retimed for transmission, but noise with a small amplitude is modulated and disperses after a transmission of a certain distance, because a low-amplitude signal is simply a dispersive wave. After a transmission of a certain distance, the next modulator is installed. Thus, the soliton is controlled and the noise remains at a low level. We assume a sinusoidal modulation for the shaping function and that its extinction ratio is . For example, means an extinction ratio of 20 dB. The signal pulse shape is assumed to be a hyperbolic secant
Fig. 23. Principle of noise reduction by the use of synchronous modulation.
where
is the ASE power from each amplifier. Here
(26)
(23) is the period of the shaping function . The where transfer function for the noise, or noise reduction ratio, due is per shaping if the sinusoidally to modulated noise disperses during propagation and becomes noise again before it reaches the next modulator. The transmitted is signal power through the modulator
(24) and is assumed to be . It is necessary to where compensate for the attenuation to maintain the soliton pulse. The . The next amplifier adds noise but required excess gain is the noise is also reduced by succeeding modulators. Thus, the total amount of noise from amplifiers is written as (25)
is always less than unity so that contherefore, (20 dB modulaverges for any value. For example, gives . The MachZehnder intion) and tensity modulator has a sinusoidal transmittivity as a function of the applied voltage, and the driving voltage is sinusoidal. Therefore, the transmission function has a steeper edge than a sinusoidal modulation, which is assumed here. In an actual situation, this asymmetric modulation function can remove the noise more efficiently. Another large advantage of soliton control is that the amplican be extended to of the order of a soliton period fier spacing . Theory suggests that the pulse distortion is proportional to [62], and dispersive nonsoliton waves are increased . This result means that with increases in should be much smaller than unity to achieve ultra-long distance soliton transmission. Higher bit rate communication requires a . The ASE noise added shorter optical pulse which shortens by the EDFAs also limits the amplifier spacing. It has been stated that the EDFA gain should be less than 10 dB because [63]. excess noise grows at a rate of These difficulties can be removed through the use of soliton control which reshapes the soliton pulse and removes dispersive nonsoliton waves as well as ASE noise [64]. B. 10-Gb/s Soliton Transmission over Unlimited Distances Our experimental setup to show the effectiveness of the soliton control is shown in Fig. 24 [65]. The soliton source was a gain-switched DFB LD operating at 1.552 m, which was converted to a transform-limited 30-ps pulse with a narrowband optical filter. The soliton data were coded using a MachZehnder LN modulator with a data speed of 10 Gb/s. The fiber loop was up to 500 km long, and an EDFA was installed
380
Fig. 24.
Experimental setup for soliton transmission over unlimited distances with soliton transmission controls in the time and frequency domains.
every 50 km. The transmission fibers were DSFs with an avps/km/nm. The soliton power was erage GVD of as low as 0.65 mW, and the average soliton period was 935 km. Soliton controls in the time and frequency domains play an important role in enabling us to maintain a repeater spacing of as much as 50 km with 2430-ps solitons. A timing clock extraction circuit was installed in the loop, and the synchronous modulator was driven by the extracted sinusoidal signal, which means that the transmitted solitons control themselves every 500 km. To realize a polarization insensitive modulator, two modulators were fusion-spliced together vertically along their polarization axes. Fixed data patterns at 10 Gb/s transmitted over 50-million km and 180-million km are shown in Fig. 25(a) and (b), respectively. The data patterns in (a) and (b) were 0100111001 and 1100110011 , respectively. It is important to note that the accumulation of ASE and nonsoliton components at 0 signal is negligible. The BER characteristics under soliton control were also measured to determine whether the present soliton control technique can really preserve data without deterioration. The BER characteristics at a bit rate of 10 Gb/s after transmis, and km were the same, which sions of means that no degradation occurred at all in the data signal after transmission over 1-million km [58], [65]. After a 1-millionkm transmission, the eye was clearly open, and this suggests that GordonHaus jitter has been eliminated by soliton control. These results indicate that soliton transmission control is a powerful technique for sending data over long distances. C. 40-Gb/s and 80-Gb/s Single-Channel DM Soliton Transmissions over 10 000 km The advantage of DM soliton transmission over a conventional WDM system is that the capacity of a single-channel becomes much larger than that of the WDM system. In this section, we describe a single-channel 40-Gb/s soliton transmission over 70 000 km and an 80-Gb/s transmission over 10 000 km realized by in-line synchronous modulation [66], [67].
Fig. 25. Transmitted data patterns at 10 Gb/s over about 50180-million km. (a) 10-bit/s data over 50-million km coded as h0100111001i. (b) 10-Gb/s data over 180-million km coded as h1100110011i.
Both experiments were performed in a 250-km DSF loop, as shown in Fig. 26, where an 80-Gb/s in-line modulation scheme is given in the dotted area. The optical soliton source was a 10-GHz harmonically mode-locked erbium fiber laser at 1550.5 nm [31]. The pulsewidth and spectral width for 40 Gb/s were 5 ps and 0.50 nm, respectively and those for 80 Gb/s were 3.0 ps and 0.81 nm, respectively. PRBS The pulse was modulated at 10 Gb/s with a using an LN intensity modulator. In order to obtain a 10-GHz
381
Fig. 26. Experimental setup for 40-Gb/s single-channel soliton transmission using in-line synchronous modulation and optical filtering. For 80-Gb/s transmission, in-line modulation was replaced with the apparatus enclosed in a dotted line.
clock signal from the transmitted 40-Gb/s signal, 10-GHz soliton units were superimposed on each other with slightly different amplitudes. This technique is also useful for reducing solitonsoliton interaction. The amplifier spacing was 50 km, and the average fiber loss for the span was 12.5 dB. We used a four-segment dispersion-decreasing configuration to reduce the influence of the dispersive waves. For the 40-Gb/s experiment, the four 12.5-km long DSFs we used had GVDs of 0.24, , and ps/km/nm. We set the average GVD 0.06, at approximately 0.04 ps/km/nm. The average launched power dBm, and the peak power into the first segment was was soliton. 50 mW, which corresponded to an almost For the 80-Gb/s experiment, the average launched power was 9.5 dBm and the peak power into the first segment soliton. The was 74 mW, which corresponded to an 80-Gb/s transmission configuration is similar to that used for the 40-Gb/s, 70 000-km transmission [66], except for the polarization multiplexing technique. To achieve soliton control at 80 Gb/s based on 40-Gb/s electronics, we employed a polarization multiplexing technique. That is, a 40-Gb/s single polarization signal was polarization-multiplexed and interleaved with an 80-Gb/s TDM signal by using a 7.5-m long PANDA fiber. After a 250-km transmission through the loop, we applied in-line synchronous modulation and narrowband optical filtering, in which the filter bandwidth was 0.80 nm for 40 Gb/s and 1.1 nm for 80 Gb/s. The 40-GHz clock signal was extracted from part of the transmitted soliton pulses with an ultrahigh-speed photodetector having a 50-GHz bandwidth and a high dielectric filter. The LN modulator for soliton control
was driven by the extracted 40-GHz clock. To obtain 80-Gb/s in-line modulation, the polarization-multiplexed 80-Gb/s signal was separated into two 40-Gb/s signals with a polarization beam splitter (PBS), and single polarization soliton control was employed at 40 Gb/s. This scheme is shown in the dotted area of Fig. 26. One clock signal was applied to two orthogonal signals. The LN modulators for soliton control were driven by the extracted 40-GHz clock, where each modulator operated at a single polarization. Simultaneously, the amplitude level of the clock signal was fed back to a polarization controller (PC) to obtain maximum clock power, which ensured that the orthogonal channel was also automatically optimized. After the in-line modulation, two orthogonal 40-Gb/s signals passed through a delay unit and were reconverted to an 80-Gb/s signal with another PBS. The transmitted 40-Gb/s soliton data signal was demultiplexed into a 10-Gb/s signal by using an electroabsorption (EA) intensity modulator. The 10-GHz clock signal was extracted from the transmitted signal. Then, by adjusting the DC bias voltage and the amplitude of the 10-GHz clock signal for the EA modulator, the gate width of the modulator was set at 20 ps to extract a 10-Gb/s signal from a 40-Gb/s signal. The demultiplexed signal was detected with a 10-Gb/s optical receiver, and BER was measured. The transmitted 80-Gb/s soliton data signal was demultiplexed into a 10-Gb/s signal by using a PBS (80 Gb/s to 40 Gb/s ) and an EA modulator (40 Gb/s to 10 Gb/s). Fig. 27(a) and (b) shows the measured BERs corresponding to 40 and 80 Gb/s. In (a), the open circles indicate the BER
382
with soliton control after a transmission of 70 000 km. The filled circles show the BER without soliton control after a transmission of 4500 km, and the triangles show the BER at 0 km. When soliton control was not employed, the maximum transmission distance was 4500 km. The BER beyond 4500 km increased rapidly with an increase in the transmission distance from the accumulation of the noise component and timing jitter. In contrast, when we employed the in-line modulation scheme [58], the maximum transmission distance was greatly extended up to 70 000 km. When the received power was greater than dBm, no error appeared at a -order error-counter . setting, which indicates that the BER was less than The inset photograph shows an eye pattern after a 70 000-km transmission, in which the eye is clearly open. In Fig. 27(b), we show eight-channel 10-Gb/s signals, four of which were are vertical and four of which were from orthogonal components. The inset photograph shows one of the eye patterns, which was demultiplexed to 10 Gb/s and the figure shows a spectral profile of the 80-Gb/s soliton signal after 10 000-km transmission. The solid line with open diamonds indicates the BER at 0 km. The power penalty after the 10 000-km transmission was typically 2.5 dB, and all data fell within a power penalty difference of 1.5 dB. When the received power was dBm, no error appeared at a -order errorlarger than counter setting, which indicates that the BER was less than . Without the in-line control, the transmission distance was approximately 1750 km, but we were able to extend it to 10 000 km by using in-line modulation. These results show that an ultrahigh bit-rate well beyond a 100-Gb/s soliton can be transmitted over 10 000 km by using in-line synchronous modulation. This technique will be more effective when the single-channel bit-rate is higher. The speed of the electronics may be limited to 100 Gb/s, and this makes all optical soliton control using high-speed optical switching more important [68]. D. Suppression of Polarization-Mode Dispersion Using In-Line Synchronous Modulation The in-line synchronous modulation technique is also useful for reducing the PMD [69]. When the bit rate of a system is increased, the system suffers from not only second- and thirdorder dispersion, but also PMD for which no effective compensation technique exists. However, PMD can be sufficiently reduced by using in-line modulation because a pulse that has been statistically broadened by PMD can be reshaped and shortened with an optical gating function. PMD causes random pulse broadening without any spectral change [70]. Therefore, spectral manipulation is not effective in reducing PMD, although the center wavelength can be fixed. In contrast, in-line modulation can in principle reduce the wings of two pulses and reshape the waveform itself [58], [59], which is useful for reducing the PMD effect. The effectiveness with which PMD induced soliton broadening can be suppressed by using soliton control was recently reported [71]. However, when soliton system design is discussed via numerical simulation, it is important to evaluate not only the evolution of single pulses, but also solitonsoliton interaction and the effect of ASE noise. When the bit rate becomes high, these effects play an important
(a)
(b) Fig. 27. BER characteristics. (a) 40-Gb/s soliton data. Open circles: BER after 70 000-km transmission with soliton control. Filled circles: BER after 4500-km transmission without soliton control. Triangles: BER at 0 km. (b) 80-Gb/s TDM soliton data.
role. Here, we evaluate the transmission quality of high-speed (40 and 80 Gb/s) soliton systems using a factor [46], [47], for various PMD conditions, and we show that in-line synchronous modulation is especially useful for extending the transmission distance in the presence of large PMD.
383
The effect of PMD can be simulated by considering the fiber as a cascade of short pieces with a constant fiber length and birefringence . At each connection point between segments, the polarization state is changed randomly, and this causes statistical coupling between two orthogonal , is given by modes. According to this model, the PMD, [71]. The propagation characteristics of the two polarized modes are governed by a coupled NLSE [70]. The conditions we adopted were as follows. The initial signal was a 32-bit pseudorandom sequence, and m. The transmission distance was defined as that factor was lower than seven. The factor at which the was calculated after the signal had passed a second-order Butterworth-type lowpass electrical filter with a bandwidth equal to 65% of the bit rate. We used a random number in the polarization rotation and calculated the mean value of the transmission distance from eight computer runs under the same condition. Here, we consider 40- and 80-Gb/s RZ systems with a repeater spacing of 80 km. The loss, third-order dispersion, effective mode area and nonlinear Kerr coefficient of the fiber were m /W, 0.25 dB/km, 0.07 ps/km/nm , 50 m , and respectively. The EDFA noise figure was 6 dB. We installed a Lorentz-shaped optical filter with a bandwidth of 3 nm at every EDFA. The GVD of the fiber, , was uniform over the whole transmission line, and we calculated the transmission distance at different GVD values. The pulse was sech shaped, and the FWHM was 5 ps for 40 Gb/s and 2.5 ps for 80 Gb/s. The peak power of the soliton pulse, , was optimized to maximize the transmission distance for each GVD value. When in-line synchronous modulation was employed, an amplitude modulator was installed every few repeaters, and the bandwidth of all of the optical filters was changed from 3 nm to 1.5 nm. Fig. 28(a) and (b) show the transmission distance in the 40-Gb/s RZ system as a function of GVD. In Fig. 28(a), the to 0.6 ps/km without PMD was varied from 0.0 ps/km soliton control. When the PMD was 0.0 ps/km , the maximum ps/km/nm, transmission distance was 4700 km ( dBm). This power corresponded to that of an soliton power, which means that an optical soliton was transmitted under this condition. However, when the PMD was 0.3 ps/km , the transmission distance was reduced to 1600 km ps/km/nm, dBm), which is of the value ( without PMD. This process clearly shows that when the PMD becomes larger, even optical solitons cannot be transmitted over long distances. When we used in-line synchronous modulation, a marked increase occurred in the transmission distance, even in the presence of large PMD. In Fig. 28(b), an in-line synchronous modulator was installed every 320 or 160 km, and the PMD was set at 0.3 ps/km . The maximum transmission distance was extended to 6800 km with a modulator spacing ps/km/nm, dBm) and further of 320 km ( extended to 11 000 km with a modulator spacing of 160 km ps/km/nm, dBm). The best result was ( obtained when the optical pulse was transmitted as an optical soliton. That is, in-line synchronous modulation is effective in extending the maximum transmission distance in the presence of large PMD.
Fig. 28. Change in the transmission distance as a function of GVD. 40-Gb/s RZ system: (a) Amplitude modulator is not installed. 0.0 ps/km . 0.1 ps/km . } 0.2 ps/km . 4 0.3 ps/km . 5 0.6 ps/km . (b) Amplitude modulator is installed every 320 km or 160 km. PMD is 0.3 ps/km . : W/O AM, filter 3 nm. : AM = 320 km, filter = 1:5 nm. }: AM = 160 km, filter = 1:5 nm. 80-Gb/s RZ system: (c) Amplitude modulator is not installed. 0.0 ps/km . 0.1 ps/km . } 0.2 ps/km . 4 0.3 ps/km . 5 0.6 ps/km . (d) Amplitude modulator is installed every 160 km or 80 km. PMD is 0.3 ps/km : W/O AM, filter = 3 nm. : AM = 160 km, filter = 1:5 nm. }: AM = 80 km, filter = 1:5 nm.
Fig. 28(c) and (d) show the transmission distance in an 80-Gb/s RZ system as a function of GVD. In Fig. 28(c), we calculated the distance without soliton control and for PMDs of 0.0 to 0.6 ps/km . It can be seen that the maximum transmission distance was 1300 km for a PMD of 0.0 ps/km ( ps/km/nm, dBm), but this was reduced ( ps/km/nm, to 500 km for a PMD of 0.3 ps/km dBm), which is about 1/3 of the value without PMD. In Fig. 28(d), an amplitude modulator was installed every 160 or 80 km for a PMD of 0.3 ps/km . The maximum transmission distance was 1500 km for a modulator spacing ps/km/nm, dBm), and it of 160 km ( was extended to 6300 km for a modulator spacing of 80 km ps/km/nm, dBm). It is also important ( to note that the maximum transmission distance was achieved when the optical pulse was transmitted as an optical soliton. This result indicates that in-line synchronous modulation is very effective in extending the maximum transmission distance in the presence of large PMD, even for a bit rate of 80 Gb/s. When we apply the in-line modulation technique, it is important to note that the transmission line has the characteristics of a soliton. Even though the line is DM, the in-line modulation is still effective as long as the dispersion swing is small, ps/km/nm. However, for example, a small GVD change of
384
Fig. 29. contour maps for evaluating WDM systems: (a) no dispersion allocation and (b) large dispersion allocation. The left and that for the DM soliton is on the right.
Q map for the NRZ pulse is on the
if the dispersion swing is of the order of 1015 ps/km/nm, the nonlinearity supporting the soliton nature becomes so weak that in-line modulation is not effective for maintaining the largely DM soliton. That is, the pulse becomes almost linear. In such a case, a soliton in-line modulator, which consists of an in-line modulator, an optical filter, and nonlinear fiber to maintain the pure soliton effect, plays an important role [72]. IX. HIGH-CAPACITY HIGH-SPEED WDM SOLITON TRANSMISSION WITH DM Map Evaluation of WDM Soliton Transmission and its A. Superb Characteristics DM soliton transmission is advantageous not only for TDM, but also for WDM in realizing a large-capacity long-distance system. The biggest advantage of WDM soliton transmission is that one channel can carry a much greater capacity than one
NRZ channel, thus enabling us to reduce the number of channels. In addition, a combination of TDM and WDM is flexible. It was thought that the WDM soliton was not so advantageous because of the soliton collisions (cross-phase modulation) between different wavelengths. However, DM of the transmission line works well to reduce the soliton collision effect and even a 1-Tbit/s (20 Gb/s 50 channels) transmission over 1000 km with a span of 100 km has been reported, in which dispersion-compensated 1.3- m SMFs were used [73]. Several reports of WDM soliton transmission have been given over ultralong distances, such as 80 Gb/s (10 Gb/s eight channels) over 9000 km using the sliding filter method [74] and 160 Gb/s (20 Gb/s eight channels) over 10 000 km with the synchronous modulation method [75]. In addition, high-capacity transmission experiments have also been demonstrated without soliton control [76], [77]. The dispersion-tuned synchronous modulation technique was also reported as an alternative to delay time
385
control [78]. When the in-line modulation method is used, it is highly advantageous to increase the bit rate as it is similar to laser mode-locking at an ultrahigh repetition rate. By contrast, chirped RZ transmission experiments with a total capacity of as high as 100160 Gb/s have been successfully reported in which the single-channel capacity was 5 Gb/s [79]. Recently, 1-Tbit/s (10 Gb/s 100 channels) transmissions over 600010 000 km have also been reported [80], [81]. In this section, we first describe the condition under which the WDM DM soliton is advantageous compared with WDM NRZ transmission, with the goal of long-distance transmission. Then, we show that it is possible to send 160-Gb/s WDM soliton data (20 Gb/s eight channels) over 10 000 km by using in-line soliton control. Fig. 29 shows maps for WDM NRZ and soliton sysmap for a WDM system, the value is tems [82]. In the represented by that of the worst channel so that the best performance is given by the worst channel. Fig. 29(a) and (b) corresponds to uniform dispersion and large dispersion allocation cases, respectively. A total of ten channels exist, and the bit rate is 10 Gb/s (a pulsewidth of 25 ps). The channel separation is 1 nm. In Fig. 29(a), one span consists of an 80-km fiber with zero GVD and a 5-km fiber with an arbitrary GVD to change the average GVD. In Fig. 29(b), the 80-km fiber has a GVD of 16 ps/km/nm for the soliton and the GVD of the 5-km DCF was varied according to the average GVD. With NRZ transmission, ps/km/nm. A disperthe 80-km fiber has a normal GVD of sion slope compensation (DSC) fiber is also installed every 320 km. In both figures, the WDM map for NRZ is on the left and that for the soliton is on the right. As shown in Fig. 29(a), WDM soliton transmission has characteristics comparable with those of WDM NRZ transmission. Although severe soliton collision occurs in a fiber with a uniform GVD, it can be reduced by adopting dispersion slope compensation, that is, by making the soliton power in each channel the same. Without DCF, the soliton transmission is worse than NRZ transmission. However, as shown in Fig. 29(b), the situation is completely reversed. That is, the WDM DM soliton transmission has much better value characteristics than does the WDM NRZ transmission, which means that soliton collisions at different wavelengths are reduced because of a process of pulse broadening in DM soliton transmission. These results indicate that the DM soliton is also applicable to WDM and provides a better result than the conventional NRZ system. Fig. 30 shows the transmission distance as a function of the dispersion of the transmission fiber (first segment). The transmission system considered here is 10 Gb/s 10 channels. One span has two segments with different dispersions. In Fig. 30(a), the dispersion of the transmission fiber (first segment) is anomalous, and in Fig. 30(b), the dispersion of the transmission fiber (first segment) is normal. We installed a DCF every 80 km (every span), and a DSCF every 320 km (every four spans). The transmitted signal power and the average dispersion were optimized for each dispersion of the transmission fiber. An interesting feature can be found in Fig. 30. When the dispersion of the transmission fiber (first segment) is small, that is, 11 ps/km/nm [less than approximately 1 ps/km/nm in Fig. 30(a) or
Fig. 30. Transmission distance as a function of the dispersion of the transmission fiber. The transmission fiber dispersion is (a) anomalous and (b) normal. A DCF is installed every 80 km, and a DSCF is installed every 320 km.
less than approximately ps/km/nm in Fig. 30(b)], a larger dispersion allocation gives a longer transmission distance at approximately the same rate for the soliton and NRZ systems. The reason for this is that FWM is suppressed when the dispersion allocation becomes larger. When the dispersion is further increased [more than approximately 1 ps/km/nm in Fig. 30(a) or more than approximately 1 ps/km/nm in Fig. 30(b)], the transmission distance increases gradually with increasing dispersion allocation in the soliton system, but it becomes almost constant and independent of the dispersion allocation in the NRZ system, because the interaction between the different channels and collision effects is still reduced in the soliton system when a soliton pulse with a high level power is broadened because of the large dispersion. However, this reduction is small for the NRZ system. When the dispersion of the transmission fiber is as much as 16 ps/km/nm, which is the dispersion of standard single-mode fiber, the transmission distance of the soliton system is almost twice that of the NRZ system. These results are in good agreement with the experimental results we reported in [83]. It is important to note that the large GVD allows us to transmit at greater power and to reduce the FWM by broadening the pulse, hence, the low peak power. Therefore, we obtain a high SNR condition even in a WDM system.
386
Fig. 31.
Experimental setup for a 160-Gb/s (20 Gb/s
2 eight channels) WDM soliton transmission in a 250-km DSF loop.

for WDM use. The total transmitted power was approximately dBm. After the 250-km transmission, the eight-channel soliton signals were separated into their respective wavelengths with a WDM coupler. To keep the same GVD in each channel, we installed a DCF in each channel after a 250-km four-step dispersion-decreasing configuration. In this case, the average GVD for all channels was set at 0.12 ps/km/nm by connecting a different DCF to each channel. The DCFs we used have GVDs of 21.3, 47.0, 72.8, 98.8, 124.9, 151.1, 177.7, and 204.5 ps/nm corresponding to signals at 1552, 1553.5, 1555, 1556.5, 1558, 1559.5, 1561, and 1562.5 nm, respectively. Because the average GVD is kept at 0.12 ps/km/nm in each channel, the average power for dBm). each channel was kept at the same value ( Then, we applied in-line synchronous modulation and narrowband filtering. Three modulators controlled two channels each, and two other modulators controlled one channel each. The bandwidth of the optical filter was approximately 0.350.45 nm. To obtain polarization-insensitive synchronous modulation characteristics, two LN modulators were connected orthogonally. This modulator was driven by an extracted 20-GHz clock for each channel. Then, while keeping the soliton power at an appropriate level, the pulses were multiplexed again through the WDM coupler and fed back to the input of the loop. The eight clock signals for in-line modulation were cleanly extracted. When each channel was not well controlled, the wing of the 20-GHz clock signal had a broad and high pedestal that did not appear when the solitons were under control. This result means that even in the presence of strong interactions between
B. 160-Gb/s [(20 Gb/s 8 and (80 Gb/s 2)] WDM Soliton Transmission over 10 000 km Using In-Line Modulation We undertook a 160-Gb/s (20 Gb/s eight channels) WDM soliton transmission experiment in a 250-km DSF loop, as shown in Fig. 31. The key to successfully increasing the system performance from 100 to 160 Gb/s is the adoption of a polarization scrambler and a phase modulator at the input [75][77]. The amplifier spacing was 50 km, and the average fiber loss for one span was 12.5 dB. We used eight LDs, which were equally spaced wavelength-stabilized between 1552 and 1562.5 nm at 1.5-nm intervals. We employed a four-segment dispersion-decreasing configuration. The use of the dispersion-decreasing configuration is a powerful way of reducing collision effects and dispersive waves, although it requires a rather more complicated selection procedure than when we use fibers with uniform dispersion. Two polarization-insensitive EA modulators were used to convert the cw beams into 1113-ps optical pulses at 20 GHz. One modulator was for signals at 1552, 1555, 1558, and 1561 nm (group 1) and the other for 1553.5, 1556.5, 1559.5, and 1562.5 nm (group 2). Two LN modulators were independently used for the data PRBS at 20 Gb/s. Data signals for group coding with a 1 were modulated with a format, and those for group 2 were modulated with a format to obtain uncorrelated neighboring channels. After amplifying each pulse to the corresponding soliton power level, all signals were polarization scrambled at low speed and synchronously phase-modulated at 20 GHz, and then introduced into the 250 km loop. The EDFAs were pumped by 1.48- m LDs and had a relatively large bandwidth
387
solitons with different wavelengths, stable solitons can be transmitted over long distances. The BERs after a 10 000-km transmission are shown in Fig. 32(a). When the received power was larger than dBm, no error appeared at a -order error-counter . setting, which indicates that the BER was less than The maximum power penalty difference was 1 dB. The inset photograph shows an eye pattern after a 10 000-km transmission at 1555 nm. When we installed a polarization scrambler, we were able to increase the number of the channels, and when we also used phase modulation, we achieved a more stable transmission with a lower penalty. Fig. 32(b) shows how the BERs degrade as the propagation distance is increased. The BERs started to worsen after a transmission of 12 000 km, which indicates that even when soliton transmission control is employed in each channel, it is not possible to send WDM soliton over unlimited distances. This result is attributed to the fact that many collisions occur among different channels. However, it was possible to send a 160-Gb/s (20 Gb/s eight channels) WDM soliton at least 10 000 km. These results indicate that ultrahigh-capacity WDM transmission over 10 000 km is also feasible through the use of soliton technology. Gb/s) WDM Fig. 33 shows our setup for 160-Gb/s ( soliton transmission over 10 000 km. In Section VIII-3, we described a single-channel 80-Gb/s soliton transmission with in-line modulation. We extended the 80-Gb/s soliton to a WDM scheme [67]. This change was not difficult because we use the same synchronous modulators for two channels. The channel wavelengths were 1550.5 and 1556 nm. We employed a polarization multiplexing technique to increase the bit rate from 40 to 80 Gb/s, which enabled us to realize soliton control at 80 Gb/s based on 40-Gb/s electronics [84]. We undertook the experiment with a 250-km fiber loop, which consisted of five spans. Each span was 50 km long, and the average fiber loss per span was 12.5 dB. We used a four-segment dispersion-decreasing configuration to reduce the influence of the dispersive waves. The four 12.5-km long , and ps/km/nm DSFs had GVDs of 0.24, 0.06, at 1550.5 nm and 0.65, 0.47, 0.37, and 0.31 ps/km/nm at 1556 nm. We set the average GVD at approximately 0.04 ps/km/nm by using DCF. The average launched powers at 1550.5 and and dBm, respectively, and the 1556 nm were corresponding peak powers into the first segment at each wavelength were 74 and 83 mW, respectively. These values also and solitons, respectively. correspond to We used two 10-GHz harmonically mode-locked erbium fiber lasers operating at 1550.5 and 1556 nm. The pulsewidth and spectral widths were 3.0 ps and 0.81 nm, respectively in both lasers. The output pulses were simulPRBS using taneously modulated at 10 Gb/s, with a an LN modulator. Then, a PLC was used to optically multiplex the 10-Gb/s signals into a 40-Gb/s data train. After that, by injecting the two 40-Gb/s signals in a direction 45 from the axis of the PANDA fiber and by using a group delay difference between the two axes, the 40-Gb/s single polarization signals were converted into two 80-Gb/s TDM signals with different wavelengths. Because the group delay difference between the two axes of the PANDA fiber at 1550.5 nm is the same as
Fig. 32. BER characteristics. (a) BER after a 10 000-km transmission. The inset photographs show the eye patterns after a 10 000-km transmission at 1555 nm. (b) Change in the BERs as a function of the transmission distance.
that at 1556 nm, we used the same PANDA fiber we used in a previous study. After a 250-km transmission through the loop, we applied in-line modulation to each wavelength and optical filtering. The filter bandwidth was 1.1 nm. First, we installed a WDM delay unit to synchronize the timing between the two channels at each modulator. Each delay unit is capable of precisely tuning the pulse peak timing. Then, we coupled two channels into a PBS. The PBS separated the polarization-multiplexed 80-Gb/s signal into two 40-Gb/s signals and single polarization soliton
388
Fig. 33.
Experimental setup for 160-Gb/s (80 Gb/s
2 two channels) soliton transmission using in-line modulation.

C. 640-Gb/s (40 Gb/s 16) Dense WDM (DWDM), DM Soliton Transmission over 1000 km Using SMFs The approaches to WDM soliton transmission described so far have focused on ultralong transmission using in-line modulation. In this section, we describe ultrahigh-capacity DM soliton transmission over 1000 km, in which a SMF was dispersion compensated by using reverse dispersion fiber (RDF). RDF is a type of DCF that can compensate for anomalous dispersion with a ratio of approximately 1 : 1 [85]. Our experimental setup for 640-Gb/s (40 Gb/s 16 channels) WDM transmission is shown in Fig. 35. The optical sources LDs that were equally were 16 wavelength-stabilized spaced from 1550.92 to 1563.05 nm, with a spacing of 100 GHz. A bit rate of 40 Gb/s and a frequency spacing of 100 GHz gives a spectral efficiency of as large as 0.4 bit/Hz, which is already well within the dense WDM (DWDM) transmission region. The signals were separated into two groups (group 1: 1550.921562.23 nm, group 2: 1551.721563.05 nm). After combining the group 1 signals, an EA modulator was used to convert the group 1 signal into 1012-ps optical pulses at 20 GHz. The same pulse generation process was applied to group 2. Two LN modulators were independently used for data PRBS at 20 Gb/s. To obtain a 40-Gb/s coding with signal, polarization and time-domain multiplexing techniques were used by employing the PMD of a PANDA fiber. To reduce the interaction between adjacent pulses, the time delay between the two polarization modes was set at 75 ps. The length of the PANDA fiber was 65 m. To achieve a stable transmission in the dispersion managed line, we applied an initial chirp of 30 ps/nm to the soliton pulses by using a high-dispersion fiber
control was employed at 40 Gb/s. Regarding the synchronous modulation, a 40-GHz clock was extracted from part of the 1550.5-nm transmitted soliton pulses by using a photodetector dielectric filter. The with a 50-GHz bandwidth and a high clock signal was then supplied to two 40-Gb/s orthogonal signals in both channels. Each single-polarization LN modulator for soliton control driven by the extracted 40 GHz had the same modulation performance for both 1550.5 and 1556 nm. Simultaneously, the amplitude level of the clock signal at 1550.5 nm was fed back to a PC to obtain maximum clock power, which ensured that the orthogonal channel was also automatically optimized. The clock signal at 1556.0 nm was also extracted and fed back to another PC. After the in-line modulation, two orthogonal 40-Gb/s signals at 1550.5 and 1556 nm were reconverted to two 80-Gb/s signals with another PBS. Fig. 34(a) and (b) show the measured BERs. We measured eight channel 10-Gb/s signals for each channel, four of which were vertical and four of which were from orthogonal components. The inset photograph shows one of the eye patterns, which was demultiplexed to 10 Gb/s after the 10 000-km transmission. The solid line with open diamonds indicates the BER at 0 km. The power penalty after the 10 000-km transmission was typically 3.5 dB, and all data fell within a power penalty difference of 2.0 dB. When the received power was larger than dBm, no error appeared at an error counter setting of , which indicates that the BER was less than . Without the in-line control, the transmission distance was approximately 1500 km, but we were able to extend it to 10 000 km by using in-line modulation. This result shows that an ultrahigh bit-rate TDM/WDM soliton can be transmitted over 10 000 km by using in-line synchronous modulation.
389
The transmitted WDM signal was demultiplexed into each wavelength component by using an arrayed-wavelength grating (AWG) demultiplexer. The transmission bandwidth and channel crosstalk of each AWG channel were 0.4 nm and less than dB, respectively. The 40-Gb/s signal was demultiplexed to 20 Gb/s using a polarization beam splitter, and then the 20-Gb/s signal was optically demultiplexed to 10 Gb/s using an EA modulator. The spectra of 16 channels at the input and after a transmission over 1000 km are shown in Fig. 36, in which the input spectrum is shown in (a) and that after transmission in (b). Fig. 37 shows the change in the received power level at BER after 1000-km transmission for each channel. Each channel consisted of four 10-Gb/s signals, and all channels had . This result indicates that a disa BER of less than persion-free, DM single-mode fiber can transmit a high-speed high-capacity DWDM soliton signal over long distances. This result also agrees well with the results in Fig. 30(a). X. DARK SOLITON COMMUNICATION In this section, we describe dark soliton transmission experiments using novel dark soliton generation and detection schemes. The solution for a dark soliton also satisfies the NLSE given by (1), where the normalized dispersion is in the normal dispersion region; hence, the sign is minus [86], [87]. The fundamental dark soliton is a tanh-type waveform (27) A dark soliton is an antisymmetric function of time, and its . In contrast, a bright soliton is an phase shifts by at even function of time, with a constant phase across the entire pulse. Several interesting reports have been published on the generation of dark solitons or dark pulses and their propagation [88][91]. Zhao and Bourkoff presented an interesting and useful technique for generating a dark soliton pulse train, in which they used a pushpull electro-optic intensity modulator [92]. When the electrical waveform of a MachZehnder modulator is expressed by a tanh-type waveform, the modulated is expressed by amplitude (28a) for the falling edge of the driving voltage (1 to 0), and (28b) for the rising edge (0 to 1), where is the amplitude of the signal of the modulator, is the phase difference of input the MachZehnder interferometer, and are the rise and fall times of the driving voltage, and is the time of one data bit. to or The phase of the modulated wave changes from to at when . In addition, a narrow from . The width of the dip decreases when dip can be seen at and are small. An important aspect of their technique is that the phase change is automatically included in the operating principle of the pushpull intensity modulator.
Fig. 34. BER characteristics of the 160-Gb/s WDM soliton data signals transmitted over 10 000 km. (a) 10-Gb/s eight-channel BERs at 1550.5 nm. (b) 10-Gb/s eight-channel BERs at 1556 nm.
because a chirped Gaussian pulse is a steady-state pulse for a DM soliton. After passing through the high dispersion fiber, the signal pulsewidth was increased to around 1618 ps because of the fiber dispersion. The WDM signal was coupled into a 250-km long recirculating loop through a 3-dB coupler. The loop consisted of five spans of fibers and EDFAs. The amplifier spacing was 50 km, and the average fiber loss was 12.0 dB. The EDFA gain tilt was compensated for by a gain equalizing filter in the fiber loop. The transmission line consisted of a 30 km of SMF and 20 km of RDF. The average GVD was set at around 0.02 0.06 ps/km/nm between 1550 and 1565 nm, and the dispersion slope was smaller than 0.005 ps/km/nm . The average coupled power to the SMF was set at 14 dBm.
390
Fig. 35. fiber.
Experimental setup for 640-Gb/s (40 Gb/s
2 16 channels) WDM soliton transmission. The transmission line consists of a SMF and a reverse dispersion
trodes is asymmetrically biased, it is accompanied by a frequency chirp at the dip of the dark soliton. In the present experiment, the rise time of the driving voltage was approximately 50 ps, which is capable of generating a dark soliton with a dip width of 50 ps. In bright soliton transmission, zero data are expressed with a low level. However, in a dark soliton system, zero data are expressed with a high level, which means that a narrower dark soliton pulse in a one data condition provides a smaller power change at the average dark soliton power level. Therefore, direct detection of the dark soliton train does not have sufficient power to discriminate zero and one when the data changes, which degrades the BER characteristics. This situation becomes serious as the pulsewidth (dip width) becomes narrower. We have developed a new scheme for detecting a dark soliton train by employing the phase difference between the wings of a dark soliton [93]. By using the interference between 1-bitshifted dark solitons, it is possible to convert the dark soliton signal to a conventional NRZ data signal. This process is shown as a dark-to-NRZ optical converter in Fig. 38, and its operating principle is shown in Fig. 39(b). The transmitted data train is separated into two arms, ARM 1 and ARM 2, in which the dark soliton data in ARM 2 is 1-bitshifted to the ARM 1 data and finally combined. This MachZehnder operation can convert the dark soliton train into . In dark soliton propagation, an inverted NRZ data signal the phase difference of between the front and rear envelopes is preserved. Hence, 1-bitshifting of the dark soliton data signals corresponding allows complete interference; and to one dark pulse (with dip) gives zero, and the interference between the same phase, corresponding to zero dark pulse (no dip), can preserve its amplitude. Thus, a dark pulse train
However, it was not possible to generate a PR dark soliton train because a true data stream cannot be generated solely by using NRZ data pulse edges. In addition, a dark soliton train cannot be detected in the same way as a bright soliton train. The dark soliton, which can be described with a sharp dip, does not greatly change its total power, which means insufficient power change occurs for the bit change. Thus, an efficient dark soliton detection scheme has been eagerly awaited. The experimental setup for dark soliton communication is DFB shown in Fig. 38. The optical source was a 1.532- m LD. The PRBS dark soliton train is generated by using an AND and -flip flop circuits and a pushpull LN intensity modulator. The operating principle is shown in Fig. 39(a). First, we use an NRZ data pattern (DATA) and its clock data (CLOCK) from a pulse pattern generator to obtain an AND signal between them using an AND circuit. This process corresponds to data conversion from NRZ to RZ with a duty ratio of 1/2. The data stream thus obtained is shown as (AND) in Fig. 39(a). This AND data stream is then converted to or by using a -flip flop circuit, in which all rising slopes change the data from 0 to 1 or 1 to 0 depending on data (see -FLIP FLOP its original state. Here, we show OUTPUT). A pushpull LN intensity modulator is used to generate dark solitons. This modulator has two symmetric electrodes on each waveguide of a MachZehnder interferometer. Each waveguide . When the is driven out of phase with a peak voltage of driving voltage was changed from a 1 to 0 or a 0 to 1 state, a narrow dip was observed at the LN output. These dips correspond exactly to the 1 state of the DATA signal. Thus, the sign of the phase can be completely changed at a dip center. When an intensity modulator is used in which one of the elec-
391
can be transformed into an NRZ signal as shown at MZ out in Fig. 39(b) which corresponds to the inverted signal of DATA in Fig. 39(a). If the phase difference between the two arms is , the output waveform of the interferometer becomes an RZ signal with a higher duty ratio, which corresponds to DATA in Fig. 39(a). Fig. 40(a) shows a 5-Gb/s fixed dark soliton data pattern 10010110 after a transmission of 400 km. Two different kinds of eye patterns resulting from data conversions using 1-bitshifting with a MachZehnder interferometer are shown in Fig. 40(c) and (d), in which (c) corresponds to conversion from a dark soliton to a conventional NRZ signal, and (d) corresponds to conversion from a dark soliton to a modified RZ signal. We were able to obtain a modified RZ signal when an optical bias of was intentionally applied to one of the two arms. This optical bias can easily be achieved by applying mechanical stress to the arm. We have examined the usefulness of the present detection technique by measuring the BER of PRBS dark solitons. A-10 PRBS dark soliton train generated by the present Gb/s, data conversion technique was detected with this method. At a , a received power improvement of as large BER of as 5 dB has been achieved, which suggests that this technique is useful for the detection of a dark soliton train. The dark soliton train was transmitted through a DSF, which has zero dispersion around 1.55 m, and was amplified with EDFAs operating at 1.53 m. To suppress the stimulated Brillouin scattering, we applied a frequency modulation with a width of approximately 150 MHz to a DFB LD at 100 kHz. The amplifier spacing was 50 km, and the average positive GVD was 1.1 ps/km/nm. The average dark soliton period was 950 km, and the fundamental dark soliton peak power was dBm. After transmission over 1200 km, a PRBS dark soliton train was detected using the novel scheme described above. The BER thus obtained is shown in Fig. 41. We achieved completely PRBS after a transmission error-free operation with , of 1000 km. A power penalty of 2 dB at a BER of which is mainly attributed to slight deviation from an ideal dark soliton and a relatively broad dip width of 50 ps in comparison with the dip separation of 100 ps. As seen in Fig. 40(a), zero, which is described with a high level, has an amplitude fluctuation and the dark soliton is slightly asymmetric at the dark soliton dip. These factors cause the power penalty. A 1200-km dark soliton transmission was possible with a power penalty increase of 0.5 dB over that at 1000 km. The inset photo in Fig. 41 is the eye pattern at 10 Gb/s after a transmission of 1000 km. It is not as clear as conventional bright soliton eye patterns, also because of the above-mentioned fluctuations, dip asymmetry and zero crossing jitter of the LN driving voltage at 10 GHz. As a conventional pulse pattern generator does not warrant the zero crossing jitter at the bit change slopes, such eye closing occurs with dark soliton use. However, it is important to note that the eye pattern degradation between 0 and 1000 km is small. Therefore, if the slope jitter is suppressed and the rising and falling slopes become completely linear without tailing, a long-distance transmission of over 10 000 km will be possible with narrower dark solitons.
Fig. 36.
Spectral distribution of 16 channels: (a) input and (b) output.
Fig. 37. Change in the received power level at a BER of 10 transmission.
after 1000-km
392
Fig. 38. Experimental setup for 10-Gb/s dark soliton communication over 1200 km.
Fig. 40. A 5-Gb/s fixed dark soliton data pattern and its data conversion. (a) A 5 Gb/s fixed dark soliton data pattern h10010110i. (b) Eye patterns after data conversion using the present method. (c) Dark soliton to conventional NRZ signal. (d) Dark soliton to modified RZ signal.
1000 km. We have also confirmed that a dark soliton pair propagates stably over 1000 km. The inset waveform is the eye pattern PRBS dark soliton train propagated over 1000 km. of a No jitter or amplitude fluctuation occurs at the 1000-km point, which suggests that a stable dark soliton can also exist in the DM fiber system. XI. SUMMARY We have described recent progress on the DM soliton. The experiments described in this paper were carried out at NTT Ibaraki and Yokosuka Laboratories over the past decade. We have attempted to explain how we arrived the DM soliton technique after exploring the average soliton or dynamic soliton technique. The DM soliton has a large power margin and dispersion tolerance compared with conventional NRZ or RZ pulse transmission at zero GVD, because the dispersion allocation (swing) enables us to keep the transmitted pulse energy at a high level as the GVD in each fiber segment is much larger than the average GVD. That is, pulse broadening occurs because of the large GVD, so that a larger energy is needed to maintain its intensity. In other words, this is a power enhancement factor. It is also important to note that simply increasing the power is insufficient to improve the transmission characteristics. Because of the dispersion allocation, nonlinear effects, such as FWM and modulation instability, cannot evolve smoothly. Therefore, even when the input power is large, the SNR can be maintained at a high level. DM soliton transmission using a conventional 1.3- m SMF seems to be a promising method as it can improve a capacity of 600 Mbit/s to 1040 Gb/s. We described transmission experiments at a single-channel 10 Gb/s over 10 000 km and at 640-Gb/s (40 Gb/s 16 channels) DWDM over 1000 km.
Fig. 39. New coding and detection schemes for a PRBS dark soliton train. (a) Coding using AND and T -flip flop circuits. (b) Detection using a 1-bitshifting technique with a MachZehnder interferometer.
The concept of the DM soliton is also applicable to dark soliton propagation by setting the average GVD so that it is normal. Fig. 42 shows the propagation characteristics of DM is 50 km, and the transmission fiber dark solitons [4]. Here, ps/km/nm and consists of a 25-km fiber with a GVD of ps/km/nm. The average a 25-km fiber with a GVD of ps/km/nm, and the pulsewidth is 20 ps. As seen GVD is in Fig. 42, the soliton pulse maintains its original value over
393
When the bit rate of a dark soliton is increased, the pulse looks similar to a conventional RZ pulse train, whereas the phase is opposite at the center of the pulse. This nature may give us a better transmission characteristic for high-speed communication. We would like to emphasize here that the DM soliton is the best way of achieving high-speed, long-distance, single-channel transmission and densely populated WDM transmission, which will play important roles in constructing functional optical networks in the 21st century. ACKNOWLEDGMENT The authors express their thanks to Dr. M. Kawachi and Dr. K. Sato for their fruitful comments and unceasing encouragement. REFERENCES
[1] H. Kubota and M. Nakazawa, Partial soliton communication systems, Opt. Commun., vol. 87, pp. 1518, 1991. [2] M. Nakazawa and H. Kubota, Optical soliton communication in a positively and negatively dispersion-allocated optical fiber transmission line, Electron. Lett., vol. 31, pp. 216217, 1995. [3] M. Suzuki, I. Morita, N. Edagawa, S. Yamamoto, H. Taga, and S. Akiba, Reduction of GordonHaus timing jitter by periodic dispersion compensation in soliton transmission, Electron. Lett., vol. 31, pp. 20272028, 1995. [4] M. Nakazawa and H. Kubota, Construction of a dispersion-allocated soliton transmission line using conventional dispersion-shifted nonsoliton fibers, Jpn. J. Appl. Phys., vol. 34, pp. L681683, 1995. [5] M. Nakazawa, Y. Kimura, K. Suzuki, H. Kubota, T. Komukai, E. Yamada, T. Sugawa, E. Yoshida, T. Yamamoto, T. Imai, A. Sahara, O. Yamauchi, and M. Umezawa, Soliton transmission at 20 Gb/s over 2000 km in Tokyo metropolitan optical network, Electron. Lett., vol. 31, pp. 14781479, 1995. [6] N. J. Smith, F. M. Knox, N. J. Doran, K. J. Blow, and I. Bennion, Enhanced power solitons in optical fibers with periodic dispersion management, Electron. Lett., vol. 32, pp. 5455, 1996. [7] M. Nakazawa, H. Kubota, A. Sahara, and K. Tamura, Marked increase in the power margin through the use of a dispersion-allocated soliton, IEEE Photon. Technol. Lett., vol. 8, pp. 10881090, 1996. [8] S. M. J. Kelly, Characteristic sideband instability of periodically amplified average soliton, Electron. Lett., vol. 28, pp. 806807, 1992. [9] Y. Kodama, S. Kumar, and A. Maruta, Chirped nonlinear pulse propagation in a dispersion-compensated system, Opt. Lett., vol. 22, pp. 16891691, 1997. [10] I. Gabitov, E. G. Shapiro, and S. K. Turitsyn, Optical pulse dynamics in fiber links with dispersion compensation, Opt. Commun., vol. 136, pp. 317329, 1997. [11] B. A. Malomed, Pulse propagation in a nonlinear optical fiber with periodically modulated dispersion: Variational approach, Opt. Commun., vol. 136, pp. 313319, 1997. [12] T. George, Extended path-average soliton regime in highly dispersive fibers, Opt. Lett., vol. 22, pp. 679681, 1997. [13] M. Matsumoto, Theory of stretched-pulse transmission in dispersionmanaged fibers, Opt. Lett., vol. 22, pp. 12381240, Aug. 1997. [14] M. Matsumoto and H. A. Haus, Stretched-pulse optical fiber communications, IEEE Photon. Technol. Lett., vol. 9, pp. 785787, Aug. 1997. [15] E. A. Golovchenko, A. N. Pilipetskii, and C. R. Menyuk, Dispersionmanaged soliton interaction in optical fibers, Opt. Lett., vol. 2, pp. 793795, 1997. [16] A. Hasegawa, Y. Kodama, and A. Maruta, Recent progress in dispersion-managed soliton transmission technology, Opt. Fiber. Technol., vol. 3, pp. 197213, 1997. [17] K. Tamura, E. P. Ippen, and H. A. Haus, Pulse dynamics in stretchedpulse fiber lasers, Appl. Phys. Lett., vol. 67, pp. 158160, July 1995. [18] H. A. Haus, K. Tamura, L. E. Nelson, and E. P. Ippen, Stretched-pulse additive pulse mode-locking in fiber ring lasers: Theory and experiment, IEEE, J. Quantum Electron., vol. 31, no. 3, pp. 591598, Mar. 1995. [19] J. H. B. Nijhof, N. J. Doran, W. Forysiak, and F. M. Knox, Stable soliton-like propagation in dispersion managed systems with net anomalous, zero, and normal dispersion, Electron. Lett., vol. 33, p. 1063, 1997.
Fig. 41. BER characteristics of 10-Gb/s dark soliton transmissions over 0, 1000, and 1200 km versus received optical power. 0 km; 1000 km; 1200 km.
Fig. 42. Propagation characteristics of DM dark soliton. The inset shows the eye pattern of a 2 1 PRBS after a 1000-km propagation.
The spectral efficiency at the 640-Gb/s experiment reached as high as 0.4 bit/Hz. Two possible ways of realizing the potential of DM soliton transmission exist. One is TDM with soliton control, which enables us to send an ultrahigh-speed signal channel transmission over 10 000 km. In this regard, we described an 80-Gb/s transmission over 10 000 km. The other is WDM, which also enables us to construct a high-speed and high-capacity system with high-speed single channels. For Gb/s transmission over this purpose, we described an 10 000 km. The difference between conventional WDM and DM soliton WDM systems is the capacity of a signal channel. The DM soliton can offer a much higher bit rate per channel. Finally, as a different approach to soliton pulse application, we reported a 10-Gb/s dark soliton transmission over 1000 km.
394
[20] V. S. Grigoryan and C. R. Menyuk, Dispersion-managed solitons at normal average dispersion, Opt. Lett., vol. 23, pp. 609611, 1998. [21] S. K. Turitsyn and E. G. Shapiro, Dispersion-managed solitons in optical amplifier transmission systems with zero average dispersion, Opt. Lett., vol. 23, pp. 682684, 1998. [22] J. N. Kutz and S. G. Evangelides Jr., Dispersion-managed breathers with average normal dispersion, Opt. Lett., vol. 23, pp. 685687, 1998. [23] A. Berntson, N. J. Doran, W. Forysiak, and J. H. B. Nijhof, Power dependence of dispersion-managed solitons for anomalous, zero, and normal path-average dispersion, Opt. Lett., vol. 23, pp. 900902, 1998. [24] Y. Chen and H. A. Haus, Dispersion-managed solitons with net positive dispersion, Opt. Lett., vol. 23, pp. 10131015, 1998. [25] T. I. Lakoba, J. Yung, D. J. Kaup, and B. A. Malomed, Conditions for stationary pulse propagation in the strong dispersion management regime, Opt. Commun., vol. 149, pp. 366375, 1998. [26] T. I. Lakoba and D. J. Kaup, Hermite-Gaussian expansion for pulse propagation in strongly dispersion managed fibers, Phys. Rev. E, vol. 58, pp. 67286741, 1998. [27] A. Hasegawa, Ed., New Trends in Optical Transmission Systems. New York: Kluwer, 1998. [28] S. K. Turitsyn, T. Schafer, and V. K. Mezentsev, Self-similar core and oscillatory tails of a path-average chirped dispersion-managed optical pulse, Opt. Lett., vol. 23, pp. 13511353, 1998. [29] H. A. Haus and Y. Chen, Dispersion-managed solitons as nonlinear Bloch waves, J. Opt. Soc. Amer., vol. B-16, pp. 889894, 1999. [30] M. Nakazawa, E. Yoshida, E. Yamada, K. Suzuki, T. Kitoh, and M. Kawachi, 80 Gbit/s soliton data transmission over 500 km with unequal amplitude solitons for timing clock extraction, Electron. Lett., vol. 30, no. 21, pp. 17771778, 1994. [31] M. Nakazawa, E. Yoshida, and Y. Kimura, Ultrastable harmonically and regeneratively modelocked polarization-maintaining erbium fiber ring laser, Electron. Lett., vol. 30, no. 19, pp. 16031604, 1994. [32] P. L. Chu and C. Desem, Mutual interaction between solitons of unequal amplitudes in optical fiber, Electron. Lett., vol. 24, pp. 11331134, 1985. [33] K. Uchiyama, H. Takara, S. Kawanishi, T. Morioka, and M. Saruwatari, Ultrafast polarization-independent all optical switching using a polarization diversity scheme in the nonlinear optical loop mirror, Electron. Lett., vol. 28, no. 20, pp. 18641865, 1992. [34] S. G. Evangelides Jr., L. F. Mollenauer, J. P. Gordon, and N. S. Bergano, Polarization multiplexing with solitons, IEEE, J. Lightwave Technol., vol. 10, no. 1, pp. 2835, 1992. [35] M. Nakazawa, K. Suzuki, E. Yoshida, E. Yamada, T. Kitoh, and M. Kawachi, 160 Gbit/s soliton data transmission over 200 km, Electron. Lett., vol. 31, no. 7, pp. 565566, 1995. [36] M. Nakazawa, K. Kurokawa, H. Kubota, and E. Yamada, Observation of the trapping of an optical soliton by adiabatic gain narrowing and its escape, Phys. Rev. Lett., vol. 65, no. 15, pp. 18811884, 1990. [37] J. P. Gordon and H. A. Haus, Random walk of coherently amplified solitons in optical fiber transmission, Opt. Lett., vol. 11, no. 10, pp. 665667, 1986. [38] A. H. Liang, H. Toda, and A. Hasegawa, High-speed soliton transmission in dense periodic fibers, Opt. Lett., vol. 24, no. 12, pp. 799801, 1999. [39] H. Anis et al., Continuous dispersion managed fiber for very high speed soliton systems, in Proc. ECOC99, Nice, France, 1999, pp. I-230I-231. [40] A. Hasegawa and Y. Kodama, Guiding-center soliton in fibers with periodically varying dispersion, Opt. Lett., vol. 16, no. 18, pp. 13851387, 1991. [41] L. F. Mollenauer, S. G. Evangelides, and H. A. Haus, Long distance soliton propagation using lumped amplifiers and dispersion-shifted fiber, IEEE J. Lightwave Technol., vol. 9, no. 2, pp. 194197, 1991. [42] M. Nakazawa, Y. Kimura, H. Kubota, T. Komukai, E. Yamada, T. Sugawa, E. Yoshida, T. Yamamoto, T. Imai, A. Sahara, H. Nakazawa, O. Yamauchi, and M. Umezawa, Field demonstration of soliton transmission at 10 Gbit/s over 2000 km in Tokyo metropolitan optical loop network, Electron. Lett., vol. 31, no. 12, pp. 992993, 1995. [43] M. Nakazawa, K. Suzuki, and Y. Kimura, Transform-limited pulse generation in the gigahertz region from a gain-switched distributed-feedback laser diode using spectral windowing, Opt. Lett., vol. 15, no. 12, pp. 715717, 1990. [44] M. Nakazawa, A. Sahara, T. Imai, T. Yamamoto, E. Yamada, and Y. Kimura, A novel technique for measuring the group velocity dispersion of an installed ultralong fiber by using erbium-doped fiber amplifiers, Jpn. J. Appl. Phys., vol. 34, pp. L1167L1169, 1995.
[45] D. Anderson, Variational approach to nonlinear pulse propagation in optical fibers, Phys. Rev., vol. A6, pp. 31353137, 1983. [46] N. S. Bergano, F. W. Kerfoot, and C. R. Davidson, Margin measurements in optical amplifier systems, Photon. Technol. Lett., vol. 5, pp. 304306, 1993. [47] A. Sahara, H. Kubota, and M. Nakazawa, -factor contour mapping for evaluation of optical transmission systems: Soliton against NRZ against RZ pulses at zero group velocity dispersion, Electron. Lett., vol. 32, pp. 915916, 1996. [48] R. W. Tkach, A. H. Gnauck, A. R. Chraplyvy, R. M. Derosier, C. R. Giles, B. M. Nyman, G. A. Ferguson, J. W. Sulhoff, and J. L. Zyskind., One-third terabit transmission through 150 km of dispersion-managed fiber, in Proc. ECOC94, Firenze, Italy, 1994, pp. 4548. [49] E. Yamada, H. Kubota, T. Yamamoto, A. Sahara, and M. Nakazawa, 10 Gbit/s, 10 600 km dispersion-allocated soliton transmission using conventional 1.3 m single-mode fibers, Electron. Lett., vol. 33, pp. 602603, 1997. [50] J. M. Jacob and G. M. Carter, Error-free transmission of dispersion-managed solitons at 10 Gbit/s over 24 500 km without frequency sliding, Electron. Lett., vol. 33, pp. 11281129, 1997. [51] G. M. Carter, J. M. Jacob, E. A. Golovchenko, and A. N. Pillipetskii, Timing-jitter reduction for a dispersion-managed soliton system: Experimental evidence, Opt. Lett., vol. 22, pp. 513515, 1997. [52] G. M. Carter and J. M. Jacob, Dynamics of solitons in filtered dispersion-managed systems, Photon. Technol. Lett., vol. 10, pp. 546548, 1998. [53] G. M. Carter, R.-M. Mu, V. Grigoryan, C. R. Menyuk, P. Sinha, T. F. Carruthers, M. L. Dennis, and I. N. Duling III, Transmission of dispersion-managed solitons at 20 Gbit/s over 20 000 km, Electron. Lett., vol. 35, pp. 233234, 1999. [54] G. M. Carter, Dispersion-managed solitons in optical communications, in Proc. SPIE, vol. 3609, 1999, pp. 4854. [55] R. I. Laming, W. H. Loh, M. J. Cole, M. N. Zervas, K. E. Ennser, and V. Gusmeroli, Fiber gratings for dispersion compensation, in Proc. OFC97, 1997, ThA3. [56] T. Komukai and M. Nakazawa, Fabrication of nonlinear chirped fiber Bragg gratings for higher-order dispersion compensation, in Proc. OFC98, 1998, TuM2. [57] E. Yamada, T. Imai, T. Komukai, and M. Nakazawa, 10 Gbit/s soliton transmission over 2900 km using 1.3 m singlemode fibers and dispersion compensation using chirped fiber Bragg gratings, Electron. Lett., vol. 35, no. 9, pp. 728729, 1999. [58] M. Nakazawa, E. Yamada, H. Kubota, and K. Suzuki, 10 Gbit/s soliton transmission over one million kilometers, Electron. Lett., vol. 27, no. 14, pp. 12701271, 1991. [59] H. Kubota and M. Nakazawa, Soliton transmission control in time and frequency domains, IEEE J. Quantum Electron., vol. 29, no. 7, pp. 21892197, 1993. [60] M. Nakazawa, H. Kubota, E. Yamada, and K. Suzuki, Infinite-distance soliton transmission with soliton controls in time and frequency domains, Electron. Lett., vol. 28, no. 12, pp. 10991101, 1991. [61] A. Mecozzi, J. D. Moores, H. A. Haus, and Y. Lai, Soliton transmission control, Opt. Lett., vol. 16, no. 23, pp. 18411843, 1991. [62] A. Hasegawa and Y. Kodama, Guiding-center soliton, Phys. Rev. Lett., vol. 66, no. 2, pp. 161164, 1991. [63] J. P. Gordon and L. F. Mollenauer, Effects of fiber nonlinearities and amplifier spacing on ultra-long distance transmission, IEEE J. Lightwave Technol., vol. 9, no. 2, pp. 170173, 1991. [64] H. Kubota and M. Nakazawa, Soliton transmission with long amplifier spacing under soliton control, Electron. Lett., vol. 29, no. 20, pp. 17801782, 1993. [65] M. Nakazawa, K. Suzuki, E. Yamada, H. Kubota, Y. Kimura, and M. Takaya, Experimental demonstration of soliton data transmission over unlimited distances with soliton control in time and frequency domains, Electron. Lett., vol. 29, no. 9, pp. 729731, 1993. [66] K. Suzuki, H. Kubota, A. Sahara, and M. Nakazawa, 40 Gbit/s single channel optical soliton transmission over 70 000 km using in-line synchronous modulation and optical filtering, Electron. Lett., vol. 34, no. 1, pp. 9899, 1998. [67] M. Nakazawa, K. Suzuki, and H. Kubota, Single-channel 80 Gbit/s soliton transmission over 10 000 km using in-line synchronous modulation, Electron. Lett., vol. 35, no. 2, pp. 162163, 1999. [68] S. Bigo, O. Leclerc, P. Brindel, G. Vendrme, E. Desurvire, P. Doussire, and T. Ducellier, 20 Gbit/s all-optical regenerator, in Proc. OFC97, 1997, PD22.
395
[69] A. Sahara, H. Kubota, and M. Nakazawa, Ultra-high speed soliton transmission in the presence of polarization mode dispersion using in-line synchronous modulation, Electron. Lett., vol. 35, no. 1, pp. 12, 1999. [70] C. R. Menyuk, Nonlinear pulse propagation in birefringent optical fibers, IEEE. J. Quantum Electron., vol. QE-23, pp. 174176, 1987. [71] M. Matsumoto, Y. Akagi, and A. Hasegawa, Propagation of solitons in fibers with randomly varying birefringence: Effects of soliton transmission control, IEEE. J. Lightwave Technol., vol. 15, pp. 584589, 1997. [72] B. Dany et al., A transoceanic 4 40 Gbit/s system combining dispersion-managed soliton transmission and new black box in-line optical modulator, Electron. Lett., pp. 418419, 1999. [73] D. LeGuen, S. DelBurgo, M. L. Moulinard, D. Grot, M. Henry, F. Favre, 20 Gbit/s) soliton and T. George, Narrow band 1.02 Tbit/s (51 DWDM transmission over 1000 km of standard fiber with 100 km amplifier span, in Proc. OFC/IOOC99, San Diego, CA, 1999, PD4. [74] L. F. Mollenauer, P. V. Mamyshev, and M. J. Neubelt, Demonstration of soliton WDM transmission at up to 8 10 Gbit/s, error-free over transoceanic distances, in Proc. OFC96, San Jose, CA, Feb. 1996, PD 22. [75] M. Nakazawa, K. Suzuki, H. Kubota, A. Sahara, and E. Yamada, 160 Gbit/s WDM (20 Gbit/s 8 channels) soliton transmission over 10 000 km using in-line modulation and optical filtering, Electron. Lett., vol. 34, no. 1, pp. 103104, 1998. [76] M. Suzuki, I. Morita, N. Edagawa, S. Yamamoto, and S. Akiba, 20 Gbit/s-based soliton WDM transmission over transoceanic distances using periodic compensation of dispersion and its slope, in Proc. ECOC96, Oslo, Sept. 1996, Paper ThB3.4. [77] M. Suzuki, N. Edagawa, I. Morita, S. Yamamoto, and S. Akiba, Soliton-based return-to-zero transmission over transoceanic distances by periodic dispersion compensation, J. Opt. Soc. Amer., vol. B-14, pp. 29532959, 1997. [78] E. Desurvire, O. Leclerc, and O. Audouin, Synchronous in-line regeneration of wavelength-division multiplexed soliton signals in optical fibers, Opt. Lett., vol. 21, no. 4, pp. 10261028, 1996. [79] N. S. Bergano et al., Long-haul WDM transmission using channel modulation: A 160 Gb/s (32 5 Gb/s) 9,300 km demonstration, in Proc. OFC97, Dallas, TX, 1997, PDP 16. [80] T. Naito, N. Shimojoh, T. Tanaka, H. Nakamoto, M. Doi, T. Ueki, and M. Suyama, 1 Terabit/s WDM transmission over 10 000 km, in Proc. ECOC99, Nice, France, 1999, PD2-1. [81] T. Tsuritani, N. Takeda, K. Imai, K. Tanaka, A. Agata, I. Morita, H. Yamauchi, N. Edagawa, and M. Suzuki, 1 Tbit/s (100 10:7 Gbit/s) transoceanic transmission using 30 nm-wide broadband optical repeaters with Aeff-enlarged positive dispersion fiber and slope-compensating DCF, in Proc. ECOC99, Nice, France, 1999, PD2-8. [82] A. Sahara, H. Kubota, and M. Nakazawa, Comparison of the dispersion allocated WDM (10 Gbit/s 10 channels) optical soliton and NRZ systems using a Q map, Opt. Commun., vol. 160, pp. 139145, 1999. [83] M. Nakazawa, E. Yamada, H. Kubota, T. Yamamoto, and A. Sahara, Numerical and experimental comparison of soliton, RZ pulse and NRZ pulses under two-step dispersion allocation, Electron. Lett., vol. 33, no. 17, pp. 14801482, 1997. [84] M. Nakazawa, K. Suzuki, and H. Kubota, 160 Gbit/s (80 Gbit/s 2 channels) WDM soliton transmission over 10000 km using in-line synchronous modulation, Electron. Lett., vol. 35, no. 16, pp. 13581359, 1999. [85] K. Mukasa, Y. Akasaka, Y. Suzuki, and T. Kamiya, Novel network fiber to manage dispersion at 1.55 m with combination of 1.3 m zero dispersion single mode fiber, in Proc. ECOC97, Edinburgh, U.K., 1997, pp. 127130. [86] V. E. Zakharov and A. B. Shabat, Exact theory of two-dimensional selffocusing and one-dimensional self-modulation of waves in nonlinear media, Sov. Phys. JETP, vol. 34, no. 1, p. 62, 1972. Zh. Eksp. Teor. Fiz., vol. 61, pp. 118134, 1971. [87] A. Hasegawa and F. Tappert, Transmission of stationary nonlinear optical pulses in dispersive dielectric fibers. 1. Anomalous dispersion, Appl. Phys. Lett., vol. 23, no. 3, p. 142, 1973. [88] P. Emplit, J. P. Hamaide, F. Reynaud, C. Froehly, and A. Barthelmy, Picosecond steps and dark pulses through nonlinear single-mode fibers, Opt. Commun., vol. 62, no. 6, pp. 374379, 1987.
[89] D. Krkel, N. F. Halas, G. Giuliani, and D. Grischkowsky, Dark pulse propagation in optical fibers, Phys. Rev. Lett., vol. 60, no. 1, pp. 2932, 1988. [90] A. M. Weiner, J. P. Heritage, R. J. Hawkins, R. N. Thurston, E. M. Kirschner, D. E. Leaird, and N. J. Tomlinson, Experimental observation of the fundamental dark soliton in optical fibers, Phys. Rev. Lett., vol. 61, no. 21, pp. 24452448, 1988. [91] D. J. Richardson, R. P. Chamberlin, L. Dong, and D. N. Payne, Experimental demonstration of 100 GHz dark soliton generation and propagation using a dispersion decreasing fiber, Electron. Lett., vol. 30, no. 16, pp. 13261327, 1994. [92] W. Zhao and E. Bourkoff, Generation of dark solitons under a cw background using waveguide electro-optic modulators, Opt. Lett., vol. 15, no. 8, pp. 405407, 1990. [93] M. Nakazawa and K. Suzuki, Generation of a pseudorandom dark soliton data train and its coherent detection by one-bit-shifting with a MachZehnder interferometer, Electron. Lett., vol. 31, no. 13, pp. 10841085, 1995. [94] M. Nakazawa and H. Kubota, Analyzes of dispersion-allocated bright and dark solitons, Jpn. J. Appl. Phys., vol. 34, pp. L889891, 1995. [95] P. A. Andrekson, N. A. Olsson, J. R. Simpson, D. J. Digiovanni, P. A. Morton, T. Tanbun-Ek, R. A. Logan, and K. W. Wecht, 64 Gb/s all-optical demultiplexing with the nonlinear optical-loop mirror, IEEE Photon. Technol. Lett., vol. 4, no. 6, pp. 644647, 1992. [96] M. Nakazawa, K. Suzuki, and Y. Kimura, Generation and transmission of optical solitons in the gigahertz region using a directly modulated distributed-feedback laser diode, Opt. Lett., vol. 15, no. 10, pp. 588590, 1990. [97] G. Aubin, T. Montalant, J. Moulu, B. Nortier, F. Pirio, and J.-B. Thomine, Record amplifier span of 105 km in a soliton transmission experiment at 10 Gbit/s over 1Mm, Electron. Lett., vol. 31, no. 3, pp. 217219, 1995. [98] B. Dany et al., Feasibility of 16 40 Gbit/s dispersion-managed transoceanic systems with high spectral efficiency, in Proc. ECOC99, Nice, France, 1999, pp. I-152I-153. [99] A. Hasegawa and F. Tappert, Transmission of stationary nonlinear optical pulses in dispersive dielectric fibers. 2. Normal dispersion, Appl. Phys. Lett., vol. 23, no. 4, pp. 171172, 1973.
Masataka Nakazawa (M84SM92F95) was born on September 17, 1952 in Yamanashi, Japan. He received the B.S. degree from Kanazawa University, Kanazawa, Japan, in 1975, the M.S. and Ph.D. degrees from the Tokyo Institute of Technology, Tokyo, Japan, in 1977 and 1980, respectively. In 1980, he joined the Ibaraki Electrical Communication Laboratory of Nippon Telegraph and Telephone Public Corporation, Kanazawa-ken, Japan, where he has been engaged in research on the transmission characteristics of optical fibers. He was with the Picosecond Optics and Quantum Electronics Group of the Massachusetts Institute of Technology, Cambridge, MA, as a Visiting Scientist from 1984 to 1985. Since 1989, he has been the group leader of the High-Speed Nonlinear Optical Transmission Research Group at NTT Laboratories and became an NTT fellow in 1999. He is also a Visiting Professor of Tohoku University. He is the author or coauthor of more than 270 journal articles and holds more than 30 patents. His current interests are optical soliton and high speed transmission, femtosecond pulse generation, and rare-earth doped optical fiber amplifiers and their applications. Dr. Nakazawa received the Niwa Memorial Award in 1980, the Best Paper Award for Optics from the Japan Society of Applied Physics in 1984, the Sakurai Memorial Award in 1989 from the OITDA, the Electronics Letters Premium Award in 1990 from the Institute of the Electrical Engineers, the Outstanding Achievement Award of IEICE in 1994 and 1996, the Minister Award of the Science and Technology Agency in 1997, and nine other awards. He is a Fellow of OSA, and a member of the Japan Society of Applied Physics and the Laser Society of Japan.
396
Hirokazu Kubota (M98) was born in Sizuoka, Japan, in 1961. He received the B.S. and M.S. degrees in physics from Osaka University, Osaka, Japan, and the Ph.D. degree in engineering from Tokyo University, Tokyo, Japan, in 1984, 1986, and 1996, respectively. He joined the Ibaraki Electrical Communication Laboratory of NTT, Yokosuka, Japan. He has been engaged in research on ultrashort pulse generation and nonlinear (especially optical solition) pulse propagation in optical fibers. Dr. Kubota is a member of the Institute of Electronics, Information and Communication Engineers (IEICE) of Japan, the Japan Society of Applied Physics, and the Optical Society of America.
Eiichi Yamada was born in Japan in May 1964. He received the B.S. and M.S. degrees in electrical engineering from Kyoto University, Kyoto, Japan, in 1987 and 1989, respectively. He joined NTT Corporation in 1989. He has engaged in research on optical soliton transmission. Mr. Yamada is a member of the IEICE and the Japan Society of Applied Physics.
Kazunori Suzuki was born in Iwate, Japan, in 1957. He received the B.S. and M.S. degrees in applied physics from Tohoku University, Sendai, Japan, in 1980 and 1982, respectively. He joined the Ibaraki Electrical Communication Laboratory, Nippon Telegraph and Telephone Public Corporation, Tokai, Ibaraki, Japan, in 1982. Since 1987, he was engaged in research on the nonlinear optics in optical fibers, especially on the optical soliton transmission. Mr. Suzuki received the Electronics Letter Premium award in 1990. He is a member of the Optical Society of America, the Institute of Electronics, Information and Communication Engineers (IEICE), the Japan Society of Applied Physics.
Akio Sahara (M97) was born in Osaka, Japan, on July 14, 1969. He received the B.S. and M.S. degrees in electrical engineering from Kyoto University, Kyoto, Japan, in 1992 and 1994, respectively. He joined NTT, Ibaraki, Japan, in 1994, where he has been engaged in research on optical solitons. He is currently working in the NTT Optical Network Systems Laboratories. Mr. Sahara is a member of the Institute of Electronics, Information and Communication Engineers of Japan (IEICE), and the Japan Society of Applied Physics.

Soliton IEEE Transaction

Diunggah oleh

Informasi Dokumen

Hak Cipta

Format Tersedia

Bagikan dokumen Ini

Bagikan atau Tanam Dokumen

Opsi Berbagi

Apakah menurut Anda dokumen ini bermanfaat?

Apakah konten ini tidak pantas?

Hak Cipta:

Format Tersedia

Soliton IEEE Transaction

Diunggah oleh

Hak Cipta:

Format Tersedia

IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 6, NO.

Ultrahigh-Speed Long-Distance TDM and WDM Soliton Transmission Technologies

1077260X/00$10.00 2000 IEEE

Experimental setup for 80-Gb/s soliton transmission over 500 km.

NAKAZAWA et al.: TDM AND WDM SOLITON TRANSMISSION TECHNOLOGIES

NAKAZAWA et al.: TDM AND WDM SOLITON TRANSMISSION TECHNOLOGIES

NAKAZAWA et al.: TDM AND WDM SOLITON TRANSMISSION TECHNOLOGIES

NAKAZAWA et al.: TDM AND WDM SOLITON TRANSMISSION TECHNOLOGIES

following coupled equations between the pulsewidth and the chirp:

NAKAZAWA et al.: TDM AND WDM SOLITON TRANSMISSION TECHNOLOGIES

BER at 40 Gb/s over 1020 km versus received optical power.

= 12 dBm; (b) average GVD = 0 ps/km/nm, coupled peak

CONTOUR MAP AND EVALUATION OF DM SOLITON TRANSMISSION SYSTEM

NAKAZAWA et al.: TDM AND WDM SOLITON TRANSMISSION TECHNOLOGIES

NAKAZAWA et al.: TDM AND WDM SOLITON TRANSMISSION TECHNOLOGIES

Dispersion tolerance of DM soliton and NRZ systems.

NAKAZAWA et al.: TDM AND WDM SOLITON TRANSMISSION TECHNOLOGIES

Fig. 23. Principle of noise reduction by the use of synchronous modulation.

is the ASE power from each amplifier. Here

NAKAZAWA et al.: TDM AND WDM SOLITON TRANSMISSION TECHNOLOGIES

NAKAZAWA et al.: TDM AND WDM SOLITON TRANSMISSION TECHNOLOGIES

Q map for the NRZ pulse is on the

NAKAZAWA et al.: TDM AND WDM SOLITON TRANSMISSION TECHNOLOGIES

Experimental setup for a 160-Gb/s (20 Gb/s

2 eight channels) WDM soliton transmission in a 250-km DSF loop.

NAKAZAWA et al.: TDM AND WDM SOLITON TRANSMISSION TECHNOLOGIES

Experimental setup for 160-Gb/s (80 Gb/s

2 two channels) soliton transmission using in-line modulation.

NAKAZAWA et al.: TDM AND WDM SOLITON TRANSMISSION TECHNOLOGIES

Fig. 35. fiber.

Experimental setup for 640-Gb/s (40 Gb/s

NAKAZAWA et al.: TDM AND WDM SOLITON TRANSMISSION TECHNOLOGIES

Spectral distribution of 16 channels: (a) input and (b) output.

Fig. 37. Change in the received power level at a BER of 10 transmission.

NAKAZAWA et al.: TDM AND WDM SOLITON TRANSMISSION TECHNOLOGIES

NAKAZAWA et al.: TDM AND WDM SOLITON TRANSMISSION TECHNOLOGIES

Anda mungkin juga menyukai