JP2010079246A - Optical waveform regenerator and optical waveform regenerating method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an optical waveform regenerator and an optical waveform regenerating method that achieve optical 3R regeneration with simple constitution. <P>SOLUTION: The optical waveform regenerator includes a pump light generating means of generating pump light having the same repeat frequency as the bit rate of input signal light and a center wavelength different from the center wavelength of the signal light, varying in intensity with time, and synchronized with the signal light, a signal mixing means of mixing the pump light and signal light with each other, and an optical waveform regenerating means of inputting the signal light and pump light mixed by the signal mixing means to a nonlinear medium and generating a nonlinear phenomenon between the pump signal and signal light to perform optical waveform regeneration of the signal light. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to an optical waveform regenerator and an optical waveform reproducing method using a nonlinear phenomenon of a nonlinear medium used in an optical fiber communication system.

  Conventionally, in an optical fiber communication system, when an optical signal is transmitted over a long distance, spontaneous emission (ASE) noise generated when the optical signal is optically amplified by an optical fiber amplifier provided in the middle of the optical fiber transmission path. Due to factors such as accumulation, dispersion characteristics of the optical fiber, and nonlinear characteristics of the optical fiber, the waveform of the optical signal after transmission is distorted, leading to deterioration in transmission quality (increase in bit error rate). This deterioration in transmission quality is particularly noticeable when the transmission bit rate of the optical signal is 40 Gbit / s or higher.

  In recent years, in order to transmit high-speed optical signals with high transmission quality over a long distance and to construct an optical communication system flexibly, optical 3R regeneration for restoring the optical quality is extremely effective, and this optical 3R regeneration technology has been studied. Yes. In particular, studies on all-optical 3R reproduction in which an optical signal is reproduced as light is underway. Optical 3R regeneration technology includes Reamplification that optically amplifies an attenuated optical signal after long-distance transmission, Reshaping that shapes the waveform by suppressing noise and fluctuation of the optical signal, and temporal fluctuation (timing jitter) of the optical signal This is a technique for realizing Retiming for correction. Further, among optical 3R regeneration technologies, a technology that realizes only Reamplification and Reshaping excluding Retiming is referred to as an optical 2R regeneration technology.

  In Non-Patent Document 1, in order to configure an optical 2R regenerator, functions of Reamplification and Reshaping are realized using an optical amplifier, a nonlinear optical fiber, and an optical band-pass fiber (OBPF).

  In Non-Patent Documents 2 and 3, the signal light and the pump light are mixed and incident on an optical fiber for generating a nonlinear effect, and based on the saturation of the gain generated through the nonlinear effect generated in the optical fiber, An optical 2R regenerator is realized by stabilizing the amplitude fluctuation.

  In Non-Patent Document 4, a short pulse light source such as a mode-locked laser is driven in synchronization with a clock extracted from a signal with respect to a signal having a bit rate of 160 Gbit / s, and the generated short pulse clock light and signal light are Retiming is realized by performing optical AND operation (optical AND gate) operation, and then the method of Non-Patent Document 1 is adopted to realize Resharping for a signal with a distorted waveform, and as an optical 3R regenerator as a whole It is functioning. By using such a method, it is possible to realize a high-speed optical 3R operation with almost no limitation on the operation speed without directly applying electrical signal processing to the ultrahigh-speed optical signal.

  In Non-Patent Document 5, pumping light and signal light are incident on a silicon chip, and using the nonlinear effect generated there, the functions of Reshaping and Retiming are realized with separate configurations.

  The configuration of a conventional optical waveform regenerator 900 is shown in FIG. In the optical waveform regenerator 900, devices having a function for realizing Retiming in the Retiming unit include, for example, an optical demultiplexer 901, a clock extractor 910, a short pulse light source 920, a signal mixing unit 903, and a nonlinear optical fiber. It comprises an optical AND operator comprising a non-linear medium 904a and a band pass optical filter 905a. Next, in order to realize Reamplification and Reshaping based on the technique disclosed in Non-Patent Document 1 in the subsequent stage, for example, optical 2R regeneration using a nonlinear medium 904b such as a nonlinear optical fiber and a bandpass optical filter 905b An optical 3R regenerator is realized by combining them.

International Publication No. 03/104886

P.V.Mamyshev, “All-Optical Data Regeneration Based On Sel-Phase Modulation Effect” ECOC1998, p.475. K. Inoue, “Optical level equalization based on gain saturation in fiber optical parametric amplifier,” Electron. Lett., Vol. 36, p. 1016 (2000). E. Ciaramella, “All-Optical Signal Reshaping via Four-Wave Mixing in Optical Fibers,” Photon. Technol. Lett., Vol.12, p.849 (2000). S. Watanabe et al., “160 Gbit / s Optical 3R-Regenerator in a Fiber Transmission Experiment” OFC2003, PD16. R. Salem et al., “Signal regeneration using low-power four-wave mixing on silicon chip” Nature Photon., Vol.2, p.35 (2008).

  As described above, in the conventional configuration, there is no known method for realizing Reshaping and Retiming at the same time, and it is necessary to separately prepare devices for realizing the respective functions. The score, size, and cost increased.

  Accordingly, the present invention has been made to solve these problems, and provides an optical waveform regenerator and an optical waveform regenerating method for realizing optical 3R reproduction with a simple configuration with reduced number of parts, size, and cost. The purpose is that.

  In order to solve the above-described problems and achieve the object, the optical waveform regenerator according to the present invention has the same repetition frequency as the bit rate of the input signal light and a center wavelength different from the center wavelength of the signal light, Pump light generating means for generating pump light whose intensity changes with time and synchronized with the signal light, signal mixing means for mixing the pump light and the signal light, and mixing by the signal mixing means An optical waveform reproducing means for inputting signal light and pump light to a nonlinear medium and generating a nonlinear phenomenon between the pump light and the signal light to reproduce an optical waveform of the signal light. And

  The optical waveform reproduction method according to the present invention has the same repetition frequency as the bit rate of the input signal light and a center wavelength different from the center wavelength of the signal light, the intensity varies with time, and the signal light A pump light generation step for generating pump light synchronized with the signal, a signal mixing step for mixing the pump light and the signal light, and the signal light and the pump light mixed in the signal mixing step are incident on the nonlinear medium And an optical waveform reproduction step of reproducing an optical waveform of the signal light by generating a nonlinear phenomenon between the pump light and the signal light.

  According to the present invention, there is an effect that optical 3R reproduction can be realized with a simple configuration.

FIG. 1 is a diagram illustrating a configuration of an optical waveform regenerator according to an embodiment. FIG. 2 is a diagram for explaining how FWM occurs and the spectrum changes when pump light and signal light are continuous light. FIG. 3 is a diagram for explaining the principle of Retiming operation. In FIG. 4, when the pump light is beat light, the intensity of the beat light is indicated by a broken line on the left vertical axis, and the instantaneous frequency change amount of the signal light is indicated by a solid line on the right vertical axis. It is a figure. FIG. 5 is a diagram illustrating calculation results of P _pump (x), P _signal (x), and Piddler (x) when the value of x changes. FIG. 6 is a diagram showing the results of calculating P _signal (α), which is the power of the output signal light, for various values of φp. FIG. 7 is a diagram illustrating a result of normalizing the vertical axis and the horizontal axis of the result illustrated in FIG. 6. FIG. 8 is a diagram illustrating a result of calculating P _signal (α) and P _idler (α) when φp = 1.5 [rad]. FIG. 9 is a diagram showing calculated values (points) and theoretical values (lines) by numerical analysis of the peak power of pump light, signal light, and primary idler light output when the fiber length is L [km]. It is. FIG. 10 is a diagram illustrating the peak power value of the signal light output when the fiber length is L [km] when the initial value of the signal light peak power is changed. FIG. 11 is a diagram illustrating calculation results of the peak time position (solid line) and the primary moment (dotted line). FIG. 12 is a diagram showing the intensity-time waveform of the signal light at the time of fiber incidence (L = 0 km) and when the fiber length L = 0.5 km by a dotted line and a solid line, respectively. FIG. 13 is a diagram illustrating values of σ _P / μ _P and <t _m ² > ^1/2 with respect to the fiber length L. FIG. 14 is a diagram illustrating calculation results of power fluctuation, timing fluctuation, and Q value with respect to the fiber length L. FIG. 15 is a diagram illustrating intensity time waveform eye patterns of (a) input signal light and (b) signal light output under the condition that the Q value is maximized. FIG. 16 is a diagram illustrating a calculation result of the Q value obtained when the peak power of the synchronized beat light that is the pump light is changed from the design value. FIG. 17 is a diagram illustrating a calculation result of the Q value obtained when the peak power average value of the signal light is changed from 77.0 mW (about 18.9 dBm). FIG. 18 is a diagram illustrating the Q value of the output signal light with respect to the fiber length L when different shapes of pump light are used. FIG. 19 is a diagram showing an intensity time waveform eye pattern of signal light output at the fiber length giving the maximum Q value in FIG. FIG. 20 is a diagram showing intensity time waveforms when signal light is cut out by OBPF after the parametric process for each of the three types of pump shapes. FIG. 21 is a diagram illustrating a result when the OBPF is replaced with an ideal filter having a pass width of 2 THz in the case illustrated in FIG. 20. FIG. 22 is an enlarged view showing only the pulses whose specified times are 0 ps and 6.25 ps in FIG. FIG. 23 is an enlarged view showing only the pulses whose specified times are 0 ps and 6.25 ps in FIG. FIG. 24 is a diagram showing an eye pattern after transmitting a 3-span optical transmission line when RZ-OOK is used as a modulation method in a numerical simulation. FIG. 25 is a diagram showing an eye pattern after transmitting a 3-span optical transmission line when RZ-DPSK is used as a modulation method in a numerical simulation. FIG. 26 is a diagram showing an eye pattern of the reproduced RZ-OOK signal corresponding to FIG. FIG. 27 is a diagram showing an eye pattern of the regenerated RZ-DPSK signal corresponding to FIG. FIG. 28 is a diagram illustrating a configuration of a signal light generator that generates an OTDM signal of 160 Gbit / s. FIG. 29 is a diagram illustrating the configuration of the optical waveform regenerator according to the embodiment. FIG. 30 is a diagram illustrating eye pattern measurement results of an OTDM signal before and after transmitting a three-stage signal quality degradation unit. FIG. 31 is a diagram illustrating a result of measuring the waveform of the signal light output from the bandpass optical filter with an optical sampling oscilloscope while changing the delay amount (Δt) of the optical delay device. FIG. 32 is a diagram showing a result of measuring the waveform of idler signal light output from the bandpass optical filter with an optical sampling oscilloscope while changing the delay amount (Δt) of the optical delay device. FIG. 33 is a diagram showing an optical spectrum at the input or output of the HNLF. FIG. 34 is a diagram illustrating an evaluation result of timing jitter with respect to the mixture ratio. FIG. 35 is a diagram showing evaluation results of amplitude jitter and signal light peak power with respect to the mixing ratio. FIG. 36 is a diagram showing a signal light waveform and an eye pattern at a point where the peak power has the maximum value. FIG. 37 is a diagram showing an optical spectrum at a point where the peak power has the maximum value. FIG. 38 is a diagram showing a signal light waveform and an eye pattern when the output power of the EDFA is 27 dBm. FIG. 39 is a diagram illustrating the peak power value of idler light output when the fiber length is L [km] when the initial value of the signal light peak power is changed. FIG. 40 is a diagram illustrating a calculation result of the peak time position and first moment of idler light. FIG. 41 is a diagram showing the intensity time waveform of idler light at L = 0.5 km and the waveform of signal light at L = 0 km. FIG. 42 is a diagram illustrating values of σ _P / μ _P and <t _m ² > ^1/2 with respect to the fiber length L for idler light. FIG. 43 is a diagram illustrating the Q value of idler light with respect to the fiber length L when different shapes of pump light are used. FIG. 44 is a diagram showing an intensity time waveform eye pattern of idler light output when the fiber length L = 0.52 km. FIG. 45 is a diagram showing the peak amplitude standard deviation σ _P of the output idler light with respect to the fiber length L. FIG. 46 is a diagram showing the peak time position standard deviation σ _T of the output idler light with respect to the fiber length L. FIG. 47 is a diagram for explaining the configuration of a conventional optical 3R regenerator.

  Hereinafter, a configuration of an optical waveform regenerator and an optical waveform reproducing method thereof according to an embodiment of the present invention will be described in detail with reference to the drawings. Note that the present invention is not limited to the embodiments. In addition, components having the same function are denoted by the same reference numerals for simplification of illustration and description.

  An optical waveform regenerator according to an embodiment of the present invention will be described below with reference to FIG. FIG. 1 is a diagram illustrating the configuration of an optical waveform regenerator.

  As shown in FIG. 1, the optical waveform regenerator 100 includes a signal light based on an optical demultiplexer 101 such as an optical coupler that demultiplexes signal light and one signal light demultiplexed by the optical demultiplexer 101. A beat light generator that generates a beat light that is a pump light that has the same repetition frequency as the bit rate and a center wavelength different from the center wavelength of the signal light, the intensity of which changes with time, and is synchronized with the signal light 102, a signal mixing unit 103 that mixes the beat light generated by the beat light generation unit 102 and the other signal light demultiplexed by the optical demultiplexer 101, and the signal mixed by the signal mixing unit 103 is a non-linear medium 104 As an optical waveform reproducing means, a nonlinear phenomenon is generated between the beat light and the signal light to perform the optical 3R reproduction of the signal, and the signal light that has been subjected to the optical 3R reproduction by the band-pass optical filter 105 is extracted. Light 3 Mainly it consists of three elements of the reproducing unit 110. Here, the beat light is light generated by superimposing two light waves having different wavelengths (frequencies), and the state in which the intensity of the beat light changes with time is the two that make up the beat light. It is represented by a sine wave shape having the frequency difference of the light wave as a beat frequency (repetition frequency).

  Here, the sine wave-shaped beat light can also be regarded as an optical pulse train having a pulse width that is half the time slot defined by the bit rate of the signal light. Hereinafter, the time slot defined by the bit rate of the signal light is appropriately referred to as a bit slot.

  For the beat light (synchronized beat light) generation unit 102 whose intensity is synchronized with the signal light, the mechanism disclosed in Patent Document 1 may be adopted, and as another example, the beat light can be synchronized with an external clock. A mechanism that limits the band of output light obtained from a short pulse light source such as a mode-locked laser by OBPF or the like may be employed.

  The shape of the pump light need not be limited to the above beat light, and may be a pulse shape such as a Gaussian type. However, the use of beat light provides the following advantages. That is, when the signal light is an ultrahigh-speed signal having a bit rate exceeding 40 Gbit / s, it is not always easy to generate a pulse train having the same repetition frequency as that of the signal light. For example, when the bit rate of the signal light is 160 Gbit / s, the generation of the pulse-shaped pump light synchronized with the signal light has a pulse width of 3 ps or less, which is a value smaller than half of the bit slot, and the repetition frequency. Means generating a pulse train that is 160 GHz, but there is no way to generate such a pulse train directly in an electrically driven device, which is generally not easy. On the other hand, beat light having a repetition frequency of 160 GHz or higher is easily generated. For example, two laser beams whose oscillation frequencies are different by 160 GHz are multiplexed, or intensity modulation or phase modulation is applied to continuous light at a frequency that is an integral fraction of 160 GHz (for example, 40 GHz). A method of cutting out the image with an arrayed waveguide grating (AWG), a Fabry-Perot filter, or the like is conceivable, and both are easy.

  As a modulation format of the input signal light, a return-to-zero (RZ) on-off keying (OOK) method generated by a method such as optical time-division multiplexing (OTDM) is used. be able to. The RZ type is not limited to the OOK system, and the method of the present invention can be applied to a phase-shift keying (PSK) system signal.

  Here, the fact that the signal light and the beat light are temporally synchronized is explained as follows. First, a prescribed time for each pulse of an optical pulse train as signal light is defined. That is, a time at which the center (intensity center) of a certain pulse exists is set as t = 0 [ps], and from there, an integer of a pulse time interval Δt (Δt = 6.25 [ps] when the repetition frequency is 160 GHz). A time that is twice as long (expressed as m × Δt using an integer m) is defined as a specified time. If the pulse train has no timing fluctuation, the center of all the pulses exists at the specified time. Conversely, the fact that the pulse train has timing fluctuation means that the center position of each pulse is randomly deviated positively or negatively from the specified time, but becomes zero when the amount of deviation is averaged with a sufficiently large number of samples. It is. The situation where the beat light is temporally synchronized with the signal light will be described using the above definition. The signal light and the beat light are temporally synchronized. The peak value of the beat light intensity is that of the signal light. This means that it matches the specified time of each pulse in the pulse train.

Next, the signal mixing unit 103 such as an optical coupler is mixed so that the beat light and the signal light are synchronized, and then input to the optical 3R reproducing unit 110. Note that an optical amplifier such as an erbium-doped fiber amplifier (EDFA) may be disposed between the signal mixing unit 103 and the optical 3R regeneration unit 110 for optical amplification. In the optical 3R reproducing unit 110, the nonlinear medium 104 is disposed at the signal input end, and an optical phenomenon is reproduced by generating a nonlinear phenomenon between the beat light and the signal light in the nonlinear medium 104. As the nonlinear medium 104, an optical fiber having a low nonlinearity and a highly nonlinear characteristic that efficiently generates a nonlinear effect is preferable. For example, it has a highly nonlinear optical fiber (hereinafter referred to as HNLF) generally having a nonlinear constant γ of 5 [1 / W / km] or more, a perforated structure, and a material other than silica glass such as tellurite or the like. It is conceivable to use a holey optical fiber or the like in which soft glass such as lead glass or chalcogenide glass is adopted. In addition, it is also possible to use various semiconductor devices, nonlinear crystals such as LN (lithium niobate, LiNbO ₃ ), LT (lithium tantalate, LiTaO ₃ ), BGO (bismuth germanate, Bi ₁₂ GeO ₂₀ ), etc. is there.

  In the subsequent stage of the nonlinear medium 104, a band-pass optical filter 105 for extracting signal light that has been subjected to optical waveform reproduction by generating a nonlinear phenomenon is disposed. Further, by using a band-pass optical filter 105 having a different center frequency, it is possible to extract not only signal light but also idler light that is simultaneously generated when a nonlinear phenomenon is generated. The bandpass optical filter 105 may be a dielectric multilayer filter, FBG (Fiber Bragg Grating), AWG, Mach-Zehnder interferometer, or the like. Furthermore, instead of the bandpass optical filter 105, a high-pass optical filter, a low-pass optical filter, or the like can be used.

  Hereinafter, the Reshaping operation principle and the Retiming operation principle that are simultaneously realized in the present invention will be described. First, a description will be given of the signal light and the time waveform and the intensity waveform when the beat light is considered as an example of the pump light.

  The complex amplitude time waveform of the signal photoelectric field is expressed as a function of time t as shown in the following equation (1).

  Further, a time waveform related to the intensity (power) of the signal light (hereinafter referred to as an intensity time waveform) is expressed as a function of time t as shown in the following equation (2).

Next, the complex amplitude time waveform of the electric field of the beat light having a wavelength (center frequency) different from that of the signal light and having the same repetition frequency as the bit rate f _B of the signal light is expressed as time t It is expressed by the function of

  Further, the intensity time waveform of the beat light is expressed as a function of time t as shown in the following equation (4).

  Here, in Expressions (1) to (4), fp and fs represent the center frequencies of the beat light and the signal light, respectively, As (t) represents the complex envelope amplitude of the signal light that changes with time, and A Represents the peak value of the complex envelope amplitude of the beat light.

Further, regarding the bit rate f _B of the signal light, a relationship of | f _p −f _s | >> f _B is established.

  Next, with reference to FIG. 2 and FIG. 3, the Reshaping operation principle and the Retiming operation principle will be described. Here, a case will be described in which an optical fiber having high nonlinear characteristics is used as the nonlinear medium, and four-wave mixing (FWM) is used as the nonlinear phenomenon of the optical fiber.

  First, a phenomenon caused by FWM will be briefly described with reference to FIG. FIG. 2 is a diagram for explaining how FWM occurs and the spectrum changes when pump light and signal light are continuous light (continuous wave; CW). In FIG. 2, the horizontal axis represents the frequency.

  As shown in FIG. 2A, the pump light 310 having a frequency of fp (wavelength λp = c / fp, where c is the speed of light in vacuum), and a frequency of fs (wavelength λs = c / fs, where c is vacuum) Consider a signal light 320 having a medium speed of light.

  By appropriately setting the power of the pump light 310 having the frequency fp, the length of the optical fiber, and the group velocity dispersion value at the frequency fp of the pump light 310, the signal light 320 having the frequency fs is amplified by the FWM. . As a result, the power of the pump light 310 decreases and becomes the pump light 310a of FIG. 2B, and the power of the signal light 320 increases to become the signal light 320a. Further, primary idler light 330 having a frequency of 2 fp-fs and higher-order idler lights 331 and 332 described later are also generated. These processes resulting from FWM are called parametric processes, optical amplification by this parametric process is called parametric amplification, and gain obtained by signal light and idler light by this parametric amplification is called parametric gain. Normally used as FWM pump light is continuous light whose amplitude is constant in time, and the gain of signal light generated by the parametric process is usually constant in time and does not change.

  On the other hand, in this embodiment, beat light is used as pump light, and since the intensity of this pump light (beat light) changes with time, the parametric gain also depends on the intensity of pump light (beat light). Change over time. As a result, even if the signal light has timing fluctuation, a Retiming effect is obtained in which the signal light is corrected by a time-dependent parametric gain.

  Next, the Retiming operation will be described in detail with reference to FIG. FIG. 3 is a diagram for explaining how the time shift of the signal light is corrected by the gain given from the beat light (pump light) of the 160 Gbit / s signal with respect to the Retiming operation principle. The horizontal axis in FIG. 3 represents time, and the vertical axis represents the relative intensity of each light when the beak power intensity of beat light (pump light) is 1.

  A waveform 401 represents beat light having a repetition frequency of 160 GHz. Further, the time during which the amplitude intensity peak of the waveform 401 exists is defined as t = 0 [ps]. A waveform 402 is a Gaussian pulse having a time width (power half-width at half maximum: FWHM) of 2.1 ps, and the peak is 0.8 ps from the peak of the amplitude intensity of the waveform 401 (in the drawing). The signal light when the time is shifted by the symbol Ta) is shown. This deviation simulates timing fluctuations that can occur when a signal is transmitted. A waveform 403 shows the signal light after the time shift of the signal light is corrected by the gain given from the beat light (pump light) by the FWM.

A waveform 401 represents the intensity of beat light | E _p (t) | ² = 1 + cos (2πf _B t), and the intensity of the signal light is represented by G (t) = 1 + | E _p (t) | ⁴ due to the effect of FWM. Received gain from beat light. At this time, the component close to t = 0 [ps] among the components of the pulse that is signal light receives a large gain because the intensity of the beat light at that time is large, but on the other hand, the time starts from t = 0 [ps]. The gain received decreases with increasing distance. With such a time distribution of gain, the waveform of the signal light is shaped so that its peak approaches the peak of the beat light on the time axis. As a result, as indicated by the waveform 403, the deviation of the center position of the signal light from the specified time is reduced to 0.58 ps (symbol Tb in the figure). Note that idler light generated by the parametric process can also be obtained as signal light with corrected time lag.

  As described above, the parametric process generated by the sync beat light is equivalent to performing a retiming operation called sync modulation in the optical domain, which has been performed using an intensity modulator driven by an electric clock. The Retiming effect can be obtained without the upper limit of the operation speed which has been a problem.

  In this embodiment, after generating the parametric process, a band pass optical filter (OBPF) is used to obtain only signal light by removing pump light and idler light. In the retiming operation of the present embodiment, the characteristics of OBPF play an important role in addition to the above operating principle. The contents are shown below.

  When FWM is generated between the pump light and the signal light, the phase of the signal light is simultaneously modulated by cross-phase modulation (hereinafter referred to as XPM). The degree of phase modulation with respect to the signal light on the time axis is proportional to the intensity of the pump light. When the pump light is not CW and the intensity changes with time, the phase of the signal light also changes with time. As a result, the instantaneous frequency of the signal light changes in proportion to the amount obtained by multiplying the time derivative of the intensity time waveform of the pump light by -1. In FIG. 4, when the pump light is beat light, the intensity of the beat light is indicated by a broken line on the left vertical axis, and the instantaneous frequency change amount of the signal light is indicated by a solid line on the right vertical axis. FIG. The horizontal axis indicates the time when t = 0 ps is the time during which the peak of the beat light intensity exists. Among the components of the signal light, the instantaneous frequency of the portion preceding the beat light (the portion of t <0 ps when t = 0 ps is set as the center of the signal light) decreases and is delayed (also t The instantaneous frequency (> 0 ps) increases. Since the center frequency of the OBPF that cuts out the signal light matches the center frequency of the signal light, the component of the signal light that has a large change in the instantaneous frequency has a large loss when passing through the OBPF. Therefore, since the output of components having a large relative time difference from the beat light among the signal light components is suppressed by OBPF, the center of energy of the intensity time waveform of the extracted signal light is the center of the signal light (t = 0 ps). ) And a Retiming effect is obtained. Further, the effect is more remarkable as the pass band of OBPF is narrower.

  As described above, in the parametric process using beat light as a pump, the time-dependent parametric amplification, the phase shift by XPM, and the passband limitation by OBPF are combined to obtain a Retiming effect for signal light having timing fluctuation.

  Next, returning to FIG. 2, the principle of Reshaping operation will be described in detail. In the following, the case where the pump light and the signal light are CW will be described, but the same argument can be made even if the signal light has a pulse shape, for example, and the pump light has a beat light shape, for example.

  The parametric gain obtained under the condition that the input power of the pump light is predetermined and the parameters of the optical fiber do not change in the longitudinal direction, the gain increases according to the fiber length while the fiber length is short. However, as the fiber length increases, the parametric gain saturates and begins to decrease. This is qualitatively explained by the combination of the following two phenomena.

  First, when FWM is generated between the pump light 310 having the frequency fp and the signal light 320 having the frequency fs, the signal light 320 receives the gain from the pump light 310 and is amplified like the signal light 320a, and further the frequency is 2fp. Primary idler light 330 that is −fs is generated. Therefore, in order to supply energy to the signal light 320a and the primary idler light 330, the power of the pump light 310 decreases and changes to the pump light 310a. Furthermore, when the power of the pump light 310a decreases, the gain for the signal light 320a also decreases, resulting in saturation of the gain obtained by the signal light.

  Second, with the growth of the parametric process, not only the primary idler light 330 but also the higher-order idler light 331 having a frequency of 3fp-2fs, the higher-order idler light 332 having a frequency of 2fs-fp, etc. The power of the expressed higher order idler light component will also grow. At this time, the signal light 320a plays the role of pump light with respect to the higher-order idler light component, and the energy with respect to the higher-order idler light component is higher than the ratio that the signal light 320a obtains gain from the pump light 310a. The proportion of supply will increase. Therefore, as a result, the energy of the signal light 310a starts to decrease after being saturated.

  Thus, when the gain of the signal light is saturated by the FWM, there is a region where the power of the output signal light does not depend on the power of the input signal light. Therefore, by applying such a condition that becomes a saturation region, it can be used as a Reshaping to suppress fluctuations in the signal amplitude and stabilize the amplitude. At this time, since the output signal light and the newly generated primary idler light both hold the phase information of the input signal light, they can be applied not only to the OOK signal but also to the PSK signal. This method is disclosed in Non-Patent Documents 2 and 3.

  Next, the effect of the difference in the modulation format of the signal light will be described with respect to the reshaping operation. When performing a reshaping operation on an RZ-OOK signal that corresponds to “mark” or “space” of the signal “mark” or “space” of the pulse, as a reshaping function, the amplitude of the pulse representing the “mark” is set to a constant value. In addition to the stabilizing function for convergence, it is ideal to have a function for removing a noise component in a time slot representing a “space” in which no pulse originally exists.

  However, the biggest factor that degrades the signal quality is a phenomenon called beat noise between the signal light and the ASE generated by the optical amplifier or the like, and this causes the amplitude of the pulse corresponding to the “mark” to fluctuate. . Therefore, even if the noise component in the “mark”, that is, the amplitude fluctuation, is removed by ignoring the noise in the “space”, the signal quality is greatly improved. For this reason, the noise removal function in the “space” is not necessarily required depending on conditions, and a large Reshaping effect can be expected even if only the amplitude stabilization function is realized. However, as described above, it is ideal for the RZ-OOK signal to have a noise removal function for both “mark” and “space”. On the other hand, when considering RZ-PSK signal reshaping, it is necessary and sufficient to stabilize the amplitude because the “mark” and “space” are expressed by the phase of the pulse, and there are pulses in all the time slots. The Reshaping function will be realized. As a method for suppressing noise in the “space” part, a method disclosed in Non-Patent Document 1, a saturable absorber device composed of a semiconductor saturable absorber mirror (SESAM) or a carbon nanotube, or a nonlinear optical loop It is conceivable to use a device such as a mirror (NOLM). By applying these methods in addition to the waveform regenerator of the present embodiment, a complete Reshaping function is realized for an OOK signal, and at the same time, a complete optical 3R regenerating function is realized. As disclosed in Non-Patent Document 3, by using high-order idler light generated by a parametric process as an output, noise in a “space” portion can be suppressed, and this is used in this embodiment. Although it is possible to apply, the energy conversion efficiency from signal light to higher-order idler light is remarkably low, and the power of the generated higher-order idler light is small, so the signal-to-noise ratio deteriorates during subsequent optical amplification Doing so is unavoidable and undesirable.

  As described above, according to the present embodiment, the resampling effect and the retiming effect are obtained by the parametric process using the synchronized beat light as the pump light, and the reamplification by the optical amplification is performed prior to the parametric process. Thus, the optical 3R reproduction function is realized.

  Further, in the present embodiment, as shown in FIG. 1, based on the beat light generating unit 102 and the optical 3R reproducing unit 110 whose operation principle is relatively simple, Reshaping and Retiming can be realized simultaneously with a simpler configuration. Can do. Therefore, the configuration as the optical 3R regenerator can be remarkably simplified, and the cost and size can be reduced.

  Next, regarding the parametric process in the case where the pump light and the signal light are CW, the phenomenon in which the gain obtained by the signal light is saturated will be described in detail using mathematical expressions and specific numerical values, and the amplitude stabilization operation of the signal light will be described. We discuss how to derive the conditions to optimize. When the dispersibility of the optical fiber is not considered, it is possible to analytically describe how the power of each component of pump light, signal light, and all idler light changes by FWM, which will be described below.

  First, the complex envelope amplitude of the input light to the optical fiber in which the pump light and the signal light are mixed is expressed by the following equation (5):

  far. In Expression (5), when the amplitude of the pump light is A and the amplitude ratio of the signal light to the pump light is α, the pump light Ep (t) having the frequency fp and the signal light Es ( t) is expressed by the following equation (6):

  It can be described by. Therefore, the mixed input light is expressed by the following equation (7):

  Can be described. However, in the equation (7), Δf represents a frequency difference Δf = fs−fp between the signal light and the pump light. If Δf is too small, crosstalk occurs between the pump light and the signal light, and also idler light generated after the parametric process. Should. This is called “providing a guard band”.

  Next, when the dispersibility of the optical fiber is not considered, the complex envelope amplitude of the output light after propagating through the optical fiber having the nonlinear constant γ [1 / W / m] and the length L [m] is expressed by the following formula ( Like 8)

  Is described. By solving this equation (8), the amplitude of a CW component having a frequency fp + mΔf, where m is an arbitrary integer (which represents pump light, signal light, and generated idler light depending on the value of m) is obtained. Each can be obtained.

Among these, the intensity P _pump (x) of the pump light when m = 0, the intensity P _signal (x) of the signal light when m = 1, and the intensity P of the primary idler light component when m = −1. _{When idler} (x) is expressed using an _nth- order first-type Bessel function J _n , respectively, as in the following equations (9) to (11),

として得られる。ここで、ｘは、下記式（１２）のように、

As obtained. Here, x is represented by the following formula (12):

である。式（１２）中のφｐ≡γＡ^２Ｌ［ｒａｄ］は、ポンプ光のみを光ファイバに入力した際の非線形位相シフト量に相当するが、信号光の振幅安定化機構の最適化設計を行う上で、重要な指標となる量である。

It is. Φp≡γA ² L [rad] in the equation (12) corresponds to a nonlinear phase shift amount when only pump light is input to the optical fiber. However, when optimizing the signal light amplitude stabilization mechanism, This is an important index.

  The results of Equations (9) to (11) describe exactly how the power of pump light, signal light, and primary idler light develops, taking into account the growth of all higher-order idler light. It is.

FIG. 5 shows the calculation results of the respective values of P _pump (x), P _signal (x), and P _idler (x) obtained by the equations (9) to (11) when the value of x changes. Show. As calculation conditions at this time, the amplitude A of the pump light is A = 1, the amplitude ratio α of the signal light to the pump light is α = 0.5, the horizontal axis in FIG. 5 is x, and the vertical axis is the power of each component. Represents. Since the value of α is a constant, changing the value of x is the same as changing φp. Further, changing φp means changing any one or more of γ, A, and L, but the value on the vertical axis in FIG. 5 changes only when the value of A is changed.

From the results of FIG. 5, it can be seen that the powers of the signal light and idler light are saturated near x = 1.5 and x = 2, respectively. Normally, since the nonlinear constant γ of the optical fiber is fixed, the fact that P _signal (x) and P _idler (x) are saturated with respect to x means that the output of signal light or idler light with respect to a certain α power can be interpreted as the presence of pumping light power a ² or the fiber length L can be saturated.

Next, using the signal strength functions of P _signal (x) and P _idler (x) obtained by the equations (10) and (11), the operation of stabilizing the signal light amplitude based on the saturation of the parametric gain is optimized. A method of obtaining the conditions for doing so will be described.

Here, φp = γA ² L is fixed to a certain value by determining the input power A ² of the pump light, the nonlinear constant γ and the length L of the optical fiber to be used, and the output signal light and idler light after the parametric process Output power of P _signal (α) = A ² [α ² J ₀ ² (2φ _p α) + J ₁ ² (2φ _p α)] and P as a function of the amplitude ratio α of the input signal light and pump light, respectively. Description will be made using the fact that it can be written as _idler (α) = A ² [J ₁ ² (2φ _p α) + α ² J ₂ ² (2φ _p α)].

FIG. 6 shows the results of calculating P _signal (α), which is the power of the output signal light, for various values of φp. In FIG. 6, the horizontal axis represents α, and the vertical axis represents the signal light power P _signal (α).

Here, when it is considered that the amplitude of the pump light is constant, α and the amplitude value of the signal light have a one-to-one relationship, and the fluctuation of the value of α means the amplitude fluctuation of the signal light as it is. That is, if the value of dP _signal (α) / dα obtained by differentiating P _signal (α) by α in FIG. 6 is 0, the power of the output signal light is constant even if the amplitude of the input signal light fluctuates. As a result, the amplitude of the signal light after the parametric process is stabilized.

Here, the minimum α that gives dP _signal (α) / dα = 0 when α> ₀ is defined as α ₀ . From the result of FIG. 6, when φp = 0.5 [rad], α ₀ does not exist in the region of 0 <α <1, and the effect of stabilizing the amplitude cannot be obtained under this condition. Although not shown in the figure, when φp = 0.5 [rad], α ₀ = 1.35 exists under the condition of α> 1, but when the amplitude of the signal light is determined using this value, the nonlinear medium ( The input signal light power to the optical fiber becomes larger than the pump light. Since it is not preferable that nonlinear effects such as self-phase modulation occur in the signal light itself, it is desirable that the amplitude of the signal light be smaller than that of the pump light, and therefore α is preferably a small value of 1 or less.

On the other hand, φp = 1.0,1.5,2.5,3.5 [rad] When, 0 <α <dPsignal in first region (α) / dα = 0 become alpha ₀ respectively alpha ₀ = It exists as 0.768, 0.555, 0.354, 0.258. Although not shown, α ₀ = 1.02 exists when φp = 0.7 [rad]. Therefore, for each φp, α ₀ obtained by calculating Expression (10) is adopted as the amplitude ratio between the pump light and the signal light, so that the amplitude stabilization mechanism by parametric amplification can be optimized. At this time, .phi.p the larger is alpha ₀ is small, also becomes small input power of the signal light, when suppressing the non-linear effect occurring in the signal light itself desirable.

Next, regarding the calculation result of the output signal light power P _signal (α) for each φp obtained in FIG. 6, the horizontal axis is normalized input power (α / α ₀ ) ² instead of α, and the vertical axis FIG. 7 shows the result of normalizing the output signal light power of the axis with P _signal (α ₀ ) which is a peak value.

From the result of FIG. 7, for each φp, the normalized output signal power is almost the same as the normalized input signal power. In principle, this does not matter which φp is selected. This means that there is no difference in the characteristics of amplitude stabilization. Specifically, if a value that is α ₀ times the amplitude of the pump light is set as the specified value of the amplitude of the signal light, the amplitude is stabilized even if the amplitude of the signal light fluctuates due to transmission. As long as the amplitude of the previous signal light is within the range of 0.8 to 1.2 with the value of (α / α ₀ ) ² , the power of the signal light obtained as an output is 1 to 0 of the specified value The characteristic that the amplitude is stabilized within the range of .97 times (range R in the figure) is obtained equally in each case of φp.

  When actually performing the amplitude stabilization operation based on the saturation of the parametric gain, it is necessary to remove only the signal light by removing the pump light and idler light with a band-pass optical filter or the like. On the other hand, even if the pump light, the signal light, and the higher-order idler light are removed and only the primary idler light is extracted, it can be obtained as output light in which the amplitude of the input signal light is stabilized.

FIG. 8 shows the results of calculating P _signal (α) and P _idler (α) when φp = 1.5 [rad]. In FIG. 8, the horizontal axis represents α, and the vertical axis represents the power of the signal light and the primary idler light output after the parametric process. P _signal (α) is the same as that shown in FIG.

From the results of FIG. 8, the value of alpha ₀ the amplitude stabilization function is obtained for the input signal light when the removal of the primary idler light is confirmed that differs from the value of alpha ₀ about previously mentioned P _{signal (α)} it can. As an example, α ₀ is 0.736 for the output power P _idler (α) of idler light, which is different from α ₀ = 0.555 in the case of P _signal (α). Therefore, when obtaining the output signal light whose amplitude is stabilized, the optimum value of α, that is, α _0, is different for achieving amplitude stabilization depending on whether the signal light is extracted as it is or the primary idler light is extracted. Become.

  As described above, the operation of stabilizing the amplitude of the signal light by appropriately selecting the nonlinear constant γ, the length L, the amplitude A of the pump light, and the amplitude ratio α of the pump light and the signal light as described above. It becomes possible to optimize.

  The conditions obtained as described above are for the case where the pump light and the signal light are both CW, but when the pump light is beat light and the signal light is in a pulse shape as in this embodiment, etc. This is also effective in general consideration of amplitude stabilization using a parametric process. In order to confirm this, the state of the parametric process that occurs when the signal light is a Gaussian pulse and the pump light is synchronous beat light is examined by numerical analysis using equation (8).

First, consider the beat light having a peak power of P = A ² = 250 mW (about 24 dBm), a center wavelength of 1550 nm, and a repetition frequency of f _B = 160 GHz as pump light, and the complex envelope amplitude of the electric field on the time axis is expressed by the equation ( Set as 3). Further, considering the equation (1) as the complex amplitude of the signal light, the complex envelope amplitude A _S (t) is a single Gaussian pulse whose power half-value width shown in the following equation (13) is ΔT _FWHM = 2.1 ps.

  think of. Here, α is a peak amplitude ratio of the signal light to the pump light. If the difference between the center frequencies of the signal light and the pump light is Δf = fs−fp = 2 [THz], the center wavelength of the signal light is approximately 1534.1 nm. It is assumed that both the beat light and the signal light have a peak amplitude at t = 0, and there is no relative time difference between the center positions of both. On the other hand, assuming that a fiber called a highly nonlinear optical fiber is used as an optical fiber for generating a parametric process, the nonlinear constant is γ = 12 [1 / W / km], and the length is L = 0.5 [ km] and the dispersion value at a wavelength of 1550 nm is zero. Note that the use of a highly nonlinear optical fiber having a larger value than the normal transmission optical fiber as the value of γ increases the generation efficiency of the parametric process, reduces the size of the optical 3R regenerator, or requires a nonlinear medium. This is effective in reducing the value of the optical power.

When these conditions are determined, φp≡γA ² L = 1.5 [rad], so that the optimum value α of the amplitude ratio between the input signal light and the pump light to be set in order to stabilize the amplitude of the signal light ₀ is α ₀ = 0.555 as described above (more accurate value is 0.55496). Therefore, the peak power of the signal light may be (Aα ₀ ) ² = 77.0 [mW].

Under the above conditions, the input light to the optical fiber is calculated by numerical analysis of Equation (8) with E ₀ (t) = E _p (t) + E _s (t) as shown in Equation (5), The state of the parametric process when the pump light as the beat light and the signal light as the Gaussian pulse are incident on the optical fiber and propagated by the distance (fiber length) L is examined. The values of the peak powers P _pump (L), P _signal (L), and P _idler (L) of the pump light, the signal light, and the generated primary idler light are rectangles whose pass width is Δf with each frequency band as the center. It is assumed that the ideal filter is applied. In addition, in order to compare with the results when the pump light and the signal light are CW, in formulas (9), (10), and (11), A ² = 250 mW, α = 0.55496, and P _pump (L) , P _signal (L) and P _idler (L) are also calculated. These results are shown in FIG.

In FIG. 9, the horizontal axis represents the distance L in the longitudinal direction of the optical fiber, and the vertical axis represents the peak power of pump light, signal light, and primary idler light. Here, the points indicated by the triangle, white circle, and black circle symbols indicate the results obtained by directly calculating Equation (8) when the pump light is beat light and the signal light is a Gaussian single pulse. In addition, the solid line and the long and short broken lines indicate the results obtained by calculating the equations (9) to (11) assuming that the pump light and the beat light are CW. Regarding the distance L on the horizontal axis, the results are shown up to the result when the fiber length exceeds a preset value of 0.5 km and is 0.8 km. The values of the peak powers P _signal (L) and P _idler (L) of the signal light and the primary idler light with respect to the fiber length L agree with the result indicated by the symbol and the line with good accuracy. Therefore, even if the pump light is beat light and the signal light is Gaussian pulse shape as in this embodiment, the amplitude is stabilized by parametric gain saturation as in the case of CW. It was also found that the numerical effects were almost the same. However, when obtaining the optimum condition α ₀ of α in the equations (9) to (11), the peak power is used as A ² instead of the average power of the beat light, and α is applied as the peak amplitude ratio of the beat light and the signal light. Note that you have to do.

In order to clarify the effect of stabilizing the amplitude of the signal light in this embodiment, the initial value of the signal light peak power is 70% with respect to the specified value (Aα ₀ ) ² = 77.0 [mW]. , 80%, 90%, 100%, 110%, 120%, 130%, and the parametric process that occurs when the conditions are the same as above except for the initial value of the signal light peak power, Calculate according to (8). One trial in this calculation is to set the input signal light peak power at a certain ratio (for example, 70%) with respect to a specified value and obtain the result of the parametric process. FIG. 10 shows the peak power value of the signal light output when the fiber length is L [km] for each trial. In FIG. 10, the horizontal axis represents the fiber length L, the left vertical axis represents the peak power of the signal light, and the solid line represents the result when the initial value of the peak power is the specified value of 77.0 mW. Further, the other lines show the cases of the ratios of 70%, 80%, 90%, 110%, 120%, and 130% of the specified values in order from the bottom at L = 0 km. At L = 0 km when the fiber is incident, the peak power of the signal light varies greatly according to each ratio. However, in the vicinity of L = 0.5 km, which is the fiber length initially set for stabilizing the amplitude. Then, it can be seen that the peak power of the output signal light converges to a substantially constant value. From this, the effect of amplitude stabilization can be directly confirmed. In FIG. 10, the value of the standard deviation σ _P of all the trial results for the output signal light peak power when the fiber length is L is also shown on the right vertical axis. The standard deviation σ _P is 1.66 mW, which is the minimum value when the fiber length is L = 0.52 km, and is approximately 1/8 of the initial value. As a result, even if the amplitude of the input signal light fluctuates from the specified value over a wide range, it converges to a range close to a constant value around L = 0.5 km, and the amplitude fluctuation of the input signal is dramatically improved. The effect of this Embodiment is shown.

  Next, assuming that the signal light has timing fluctuations (random position shift on the time axis), the time for giving the signal light peak on the time axis (hereinafter referred to as the peak time position) is the beat. Considering the situation deviating from the peak time position of light, the verification of the signal light timing fluctuation correction effect according to the present invention and the influence of the shape of the pump light on the effect will be examined.

As the shape of the pump light, in addition to the beat light with the repetition frequency f _B = 160 GHz considered above, three types of Gaussian pulses with a half-value width of 2.5 ps or 1.5 ps are considered, and in all cases the peak power Is set to 250 mW, and the peak time position is set to 0 ps. As the signal light, a Gaussian single pulse having a power half width of 2.1 ps is considered as described above, and the peak time position is set to 0.5 ps. At this time, the peaks of the pump light and the signal light have a relative time difference of 0.5 ps. The state of the parametric process that occurs when any one of the three types of pump light and the signal light are mixed and propagated through the optical fiber under the same conditions as described above is calculated by equation (8), and the fiber length L The peak time position of the signal light with respect to and the first moment shown in the following formula (14)

  Got. The first moment means the barycentric position of the pulse energy on the time axis, and it is more important than the peak time position as an amount to quantitatively evaluate timing fluctuations when considering improvement of bit error characteristics during signal reception. It is.

  FIG. 11 shows the calculation results of the peak time position and the first moment. In FIG. 11, the peak time position is indicated by a solid line, and the first moment is indicated by a broken line. Further, in FIGS. 11A and 11B, (a), (b), and (c) respectively represent pump light as beat light, a Gaussian pulse with a power half width of 2.5 ps, and a Gaussian pulse with a power half width of 1.5 ps. The result is shown. The horizontal axis indicates the distance L of the propagated fiber, and the vertical axis indicates either the peak time position or the first moment. 11 (a), (b), and (c), the fiber length L increases, and both the peak time position and the first moment of the signal light approach 0 ps. It can be confirmed that the time lag of the signal light is compensated by the parametric process of the present invention.

  When attention is paid to the peak time position and the first moment of the signal light in FIG. 11, the peak time position indicated by the solid line is greatly approaching 0 ps as the fiber length increases in all cases (a), (b), and (c). However, the amount of change in the primary moment shown by the broken line is relatively small, and there is a difference between the two values. When the time waveform of a pulse is symmetric in time, such as a Gaussian function, the difference between the two is zero, so a large difference means that the waveform of the signal light is distorted and the quality is low. To do. In the case of (c), which is a Gaussian pulse with a power half width of 1.5 ps, the amount of change in the primary moment with respect to the fiber length L is smaller than in the cases of (a) and (b), and the fiber length L = 0. The difference between the primary moment and the peak time position at .5 km is the largest value.

  In order to confirm the waveform distortion of the signal light, the intensity time waveform of the signal light at the time of fiber incidence (L = 0 km) and when the fiber length L = 0.5 km is shown in FIG. In FIG. 12, (a), (b), and (c) correspond to (a), (b), and (c) in FIG. A broken line indicates a case where L = 0 km, and a solid line indicates a case where L = 0.5 km. From FIG. 12, the signal light having a symmetric waveform on the time axis when entering the fiber changes to an asymmetric waveform after propagating through the optical fiber by 0.5 km. In particular, in the case of (c), it is more asymmetrical than (a) and (b). Considering the purpose of improving the quality of signal light, it is not preferable that the waveform is asymmetric. Therefore, in order to correct the timing fluctuation of the signal light, in order to obtain a high quality, that is, an output signal light waveform that is symmetrical on the time axis, it is preferable to use a pulse shape with a relatively large beat light or time width.

  Next, assuming that the input signal light has both amplitude fluctuation and timing fluctuation, and when they are corrected by the parametric process, the influence of the shape of the band transmission optical filter (OBPF) for extracting the signal light is affected. Consider. Specifically, the result of changing from the ideal filter (which does not exist in reality) used in the previous studies to a Gaussian filter with three different bandwidths as an example of a realistic filter is shown. .

The optical fiber conditions are the same as in the above study. The pump light is beat light having a center wavelength of 1550 nm, a repetition frequency f _B = 160 GHz, and a peak power of 250 mW. As a condition of the signal light, the shape is a Gaussian single pulse with a time width (FWHM) of 2.1 ps, and the specified value of the peak power is 77.0 mW at which the operation of stabilizing the amplitude is optimum. The difference between the peak time positions of the signal light and the pump light at the time of incidence of the optical fiber is the initial time difference Δt, and the values of Δt are Δt = 0, 0.3, 0.5, 0.7, 0.9 ps. think of. The frequency difference between the signal light and the pump light is Δf = 2 THz. As the shape of the OBPF that cuts out the signal light after the parametric process, there are a total of four types: an ideal filter with a pass width of 2 THz and a Gaussian filter with a 3 dB bandwidth Δf _{3 dB} of 0.5, 0.3, and 0.2 THz. think of. For each condition, similarly to the case of FIG. 10, the peak power of the input signal light is changed at a ratio of 70%, 80%, 90%, 100%, 110%, 120% and 130% of the specified value, and the equation (8) is changed. To determine the parametric process in each case. First, one initial time difference Δt between the signal light and the pump light and one type of OBPF are selected, and this is set as one condition. Next, one trial in this calculation is to change the input signal light peak power at a certain ratio described above with respect to the specified value and obtain the result of the parametric process. For all trials under one condition, after calculating the peak power P of the output signal light with respect to the fiber length L and the primary moment t _m on the time axis, the standard deviation σ _{P of P} and the ratio σ _P / _{μ P} average value mu _P, to evaluate each value of the square root ^_<t m ^{2> 1/2} of mean square _{t m.} When the bandwidth of OBPF is different, the mean value μP of _P is different, but the initial value of σ _P / μ _P is 0.2 in all conditions and in all trials.

FIG. 13 is a diagram showing values of σ _P / μ _P and <t _m ² > ^1/2 with respect to the fiber length L for each condition. In FIGS. 13A, 13B, 13C, 13D, and 13E, the initial time difference between the pump light and the signal light is Δt = 0, 0.3, 0.5, 0.7. , 0.9 ps, and in each case, four types of lines are obtained by applying an ideal filter and three types of Gaussian filters having different ₃ dB bandwidths Δf and 3 dB as OBPF for extracting signal light. It is a result. In (a), since Δt = 0, <t _m ² > ^1/2 is always 0.

In FIG. 13, when attention is paid to the difference in the result due to the difference in the filter with respect to the value of <t _m ² > ^1/2 in a certain fiber length, a filter with a small bandwidth is used in all cases (b) to (e). However, the value of <t _m ² > ^1/2 is smaller, and it can be said that the effect of correcting the time position shift of the signal is higher. As described with reference to FIG. 4, this result shows that the Retiming effect is obtained by band limitation by XPM and OBPF generated with respect to signal light, and the effect is that the narrower band of OBPF is improved. Prove that. However, if the bandwidth of the OBPF is too narrow, the time width of the signal light pulse widens and intersymbol interference occurs on the time axis, which is not preferable. It is important to narrow the bandwidth within a range where no intersymbol interference occurs. It is. Further, in all cases (b) to (e), the value of <t _m ² > ^1/2 is smaller as the fiber length L is longer, and the effect of correcting the time position deviation of the signal is enhanced. It can be seen that it is better to set the fiber length longer.

On the other hand, with respect to the minimum value of σ _P / μ _P that is a measure of the degree of the amplitude stabilization effect, although there is no clear difference due to the difference in the bandwidth of OBPF, as Δt increases, σ _P / μ _P The fiber length from which the minimum value can be obtained is increased. Specifically, in the case of Δt = 0 ps in (a), as shown in the result of FIG. 10, σ _P / μ _P has a minimum value around L = 0.5 km, and the amplitude of the signal light In contrast to the original design aimed at minimizing fluctuations, when Δt = 0.9 ps in (e), L = 0.6 km or more. This means that when the initial time difference Δt between the pump light and the signal light is large, the amount of frequency shift by XPM becomes large, and the power is removed by the subsequent OBPF, so that a large gain is necessary. The fiber length until a sufficient parametric gain is obtained as a result is due to the two effects of the effect and the effect that the parametric gain is reduced because the power of the pump light is small at the time when the peak power of the signal light is given. It is thought to be longer. Therefore, in order to perform amplitude stabilization in a situation where the signal light has a large initial time difference from the pump light, the fiber length L should be set to a value larger than a value set considering only amplitude stabilization. Since this essentially increases φp = γA ² L in equation (12), instead of increasing the fiber length L, the nonlinear constant γ of the fiber and the peak power A of the pump light ² may be increased.

  In the examination so far, a single pulse is considered as the signal light, and in order to simulate the amplitude fluctuation and timing fluctuation, there are the ratio to the specified value of the amplitude and the relative time difference between the peak time positions of the pump light and the signal light. We have determined the values and analyzed how they are corrected by the parametric process. In the following, a pulse train is considered as input signal light, and when each pulse has random amplitude fluctuations and timing fluctuations, the manner in which those fluctuations are corrected by the Reshaping effect and the Retiming effect of the present embodiment is examined. A method for obtaining an optimum condition as an optical waveform regenerator and an optical waveform reproduction method under realistic conditions using numerical simulation will be described. The conditions assumed in the numerical simulation are as follows.

Assuming that beat light having a repetition frequency of 160 GHz is considered as pump light, the peak power is A ² = 250 mW (about 24 dBm), and the center wavelength is 1550 nm. At this time, the average power of the beat light is 125 mW (about 21 dBm).

Considering an in-phase pulse train having a repetition frequency of 160 GHz as input signal light, the time waveform of each pulse is a Gaussian type with a half-power width of 2.1 ps, and the center wavelength is 1534.1 nm. At this time, the frequency difference between the signal light and the pump light is 2 THz. Α ₀ = 0.555 is used as the peak amplitude ratio of the signal light pulse with respect to the beat light that is the pump light, and the peak power prescribed value μ _p of the signal light is μ _p = (Aα ₀ ) ² = 77.0 [ mW]. This prescribed value μ _p is an average value when random amplitude fluctuation is given to the signal light pulse later. Data modulation by intensity modulation or the like is not performed. Amplitude fluctuation and timing fluctuation are given to the pulse train of this input signal.

As for the amplitude fluctuation, fluctuation by a normally distributed random number is given such that the standard deviation σ _P of the signal light peak power is 7.7 mW which is 10% of the peak power average value μ _P. As for timing fluctuation, the center position (peak time position) of each pulse is randomly shifted from the predetermined time t _0, and the average value is 0 ps and the standard deviation is σ _T = 0.5 ps. Normal distributed random numbers are used.

  Under the above conditions, when the beat light and the signal light are temporally synchronized and superimposed, and input to an optical fiber having a nonlinear constant γ = 12 [1 / W / km], which is a nonlinear medium. Then, a simulation is performed to obtain the waveform of the output signal light and the values of various parameters that are propagated through the optical fiber and regenerated by the parametric process generated there. As a rule, a Gaussian shape is applied as the shape of the OBPF used to remove the pump light and idler light after the parametric process and cut out only the signal light.

First, the fluctuation of the signal light amplitude and timing and the change of the signal quality when the fiber length L of the highly nonlinear optical fiber is changed from 0 km to 0.8 km are examined by simulation. When a value obtained by dividing the standard deviation σ _P of the peak power of the pulse by the average value μ _P is adopted as a value for quantitatively evaluating the amplitude fluctuation, the initial value is σ _P / μ _P = 0.1. When the standard deviation σ _T of the pulse center position (time to give peak power) is adopted as a value for quantitatively evaluating the timing fluctuation, the initial value is σ _T = 0.5 ps. Furthermore, although a Q value often used as a parameter representing signal quality is considered, since a signal pulse train that does not perform intensity modulation applied in the OOK system is considered as a format of signal light, the Q value is defined as the signal pulse train. The statistical value at time t ₀ is defined as a value μ _P / σ _P | _{t =} t ₀ obtained by dividing the average value μ _P of the pulse intensity by the standard deviation σ _P. Note that changing the fiber length L of the highly nonlinear optical fiber corresponds to fixing α in Expressions (9) to (12) and changing only x. Therefore, if the value of x is the same, the same result can be obtained even if the peak power A ² of the pump light or the nonlinear constant γ of the optical fiber is changed instead of the fiber length L.

FIG. 14 shows the fiber length L shown on the horizontal axis and (a) σ _P representing the power fluctuation shown on the vertical axis when the bandwidth Δf _{3 dB} of the Gaussian OBPF that extracts the signal light is changed in three ways. It is a figure which shows the calculation result of the relationship with (/ _P ), (b) (sigma) _T showing timing fluctuation, and (c) value of ( _micro | _micron | mu) _P / (sigma) _P | _{t = t0} showing Q value.

From the result shown in FIG. 14A, the value of the amplitude fluctuation σ _P / μ _P whose initial value was 0.1 is when the fiber length L is about 0.6 km regardless of the bandwidth Δf _{3 dB of the} BPF. 10 and is longer than 0.5 km, which is the fiber length shown in FIG. 10 where the amplitude fluctuation is minimized when there is no timing fluctuation. As shown in FIG. 13, this reflects the result that the fiber length at which the amplitude fluctuation of the output signal light is minimized becomes longer as the timing fluctuation of the input signal light becomes larger. In addition, the amplitude fluctuation is smaller as Δf _{3 dB} is larger.

On the other hand, from the result shown in FIG. 14B, the timing fluctuation σ _T decreases from the initial value of 0.5 ps as the fiber length L increases. Here, the initial value of σ _T in the case of Δf _{3 dB} = 0.2 THz is a value different from 0.5 ps. This is due to a numerical calculation error caused by intersymbol interference, and the value of σ _T is As the length decreases, this error is eliminated. When the fiber length L is greater than 0.4 km, it can be seen that the smaller Δf _{3 dB} is, the smaller σ _T is, and the smaller the OBPF bandwidth is, the higher the timing fluctuation correction capability is.

Finally, from the results shown in FIG. 14 (c), the Q value representing the overall quality of the signal has the maximum fiber length in the vicinity of 0.6 km for any Δf _{3 dB} , and the amplitude fluctuations. Is consistent with the condition to minimize. From this, it can be seen that when the fiber length is larger than the fiber length at which the amplitude fluctuation is minimized, the amplitude fluctuation greatly increases, and this is a dominant factor and the Q value also deteriorates. In addition, the maximum Q value is obtained when Δf _{3 dB} is the smallest 0.2 THz. Of the amplitude fluctuation and the timing fluctuation, Δf _{3 dB} , the timing fluctuation affects the Q value. This is probably because of this. Therefore, it can be concluded that it is better to use OBPF with a small bandwidth in order to improve the quality of the reproduced signal.

  As described above, as a result of examination of the optimum design of the optical waveform regenerator under realistic conditions, it has been found that the fiber length should be set while reducing the filter band as much as possible. On the other hand, it is also possible to perform optimization by fixing the fiber length to a constant value and changing the peak power of the pump light and the power of the signal light to perform the same calculation. Note that the optimum condition obtained here is considered to vary depending on the amplitude fluctuation and timing fluctuation of the input signal light.

The above conditions for maximizing the Q value, that is, the signal light output when the fiber length L is 0.6 km, and the bandwidth Δf _{3 dB} of the OBPF for cutting out the signal light is 0.2 THz, and the intensity of the input signal light The time waveform eye pattern is shown in FIG. FIG. 15A shows an eye pattern of input signal light, and FIG. 15B shows an eye pattern of output signal light.

  From the result of FIG. 15, in the input signal light eye pattern of FIG. 15A, a large fluctuation (jitter) is seen in the amplitude and timing of the pulse, but when attention is paid to the output signal light eye pattern of FIG. It can be seen that fluctuations in both amplitude and timing are significantly suppressed.

  In the result of FIG. 15B, the variation in intensity is observed when the time on the horizontal axis is around ± 3 ps. The reason is that the intensity of the pump light at this time is almost zero, and the intensity of the signal light is parametric. This is because the amplitude and time fluctuations that were initially given remain as they are because they are not subjected to amplification, and thus neither increase nor decrease. As shown in another embodiment, when idler light is cut out by OBPF instead of signal light, idler light is not generated when the pump light intensity is zero, so the intensity around time ± 3 ps is It becomes zero, and there is no fluctuation in amplitude, and a more powerful waveform shaping effect and high-quality waveform output are expected.

In the results shown in FIGS. 16 and 17 shown below to calculate the Q value when the peak power average value mu _P peak power A ² and the signal light of the pump light is changed from the optimum value, respectively, the pump light and the signal light Confirm that the signal quality changes when the power of the signal deviates from the design value.

First, when the fiber length of the highly nonlinear optical fiber is 0.6 km, the average value μ _P of the peak power of the input signal light is fixed to 77.0 mW, and the standard deviation σ _P is fixed to 7.7 mW, the synchronous beat that is the pump light FIG. 16 shows the calculation result of the Q value obtained when the peak power A ^{2 of} light is changed from the design value of 250 mW (about 24 dBm) as shown on the horizontal axis.

From the results shown in FIG. 16, the peak power A ² synchronization beat light has become Q value is the maximum near the design value 24 dBm, but the change in Q value in the range of ± 0.3 dB from the design value is substantially flat Above this range, the Q value is significantly reduced. Thus, for the value of the peak power A ² synchronous optical beat, it should be set the value of the design values to possible critical. Note that secure the peak power average value mu _P of the signal light, changing the peak power A ² of the beat light is equivalent to varying both the α and x in formula (9) to (12) .

Next, the fiber length is fixed to 0.6 km, the peak power of the beat light is fixed to A ² = 250 mW (about 24 dBm), and the peak power average value μ _P of the signal light is changed from 77.0 mW (about 18.9 dBm) to the horizontal axis. FIG. 17 shows the calculation result of the Q value obtained when changing as shown. However, the ratio σ _P / μ _P between the standard deviation of the signal light peak power and the average value is fixed to 0.1, and the value of the standard deviation σ _P is also changed according to the value of the average value μ _P.

  In FIG. 17, the Q value is maximized in the vicinity of 18.9 dBm, which is the peak power average value of the signal light. From the results shown in FIG. 17, since the Q value rapidly deteriorates when the peak power average value of the signal light deviates about ± 0.3 dB or more from the specified value, the peak value of the pump light is also related to the average peak power of the input signal light As with power, you should set the value exactly as designed.

  From the above-described numerical simulation results, the effect of optical waveform reproduction based on parametric amplification generated between the beat light and the signal light is confirmed under realistic conditions, and both the Retiming effect and the Reshaping effect are included. The optimum operating conditions were confirmed. In the obtained result, white noise such as ASE noise is not considered, but the same result can be obtained even when this is taken into consideration.

  Next, the influence of the shape of the pump light on the optical waveform reproduction operation of the present invention under realistic conditions will be considered. Specifically, the results for the case of using a Gaussian antiphase pulse train shape in addition to the sine wave shape of the beat light used in the above examination as the shape of the pump light are shown. Note that the anti-phase pulse train is a pulse train in which adjacent pulses are phase-shifted by π radians.

  FIG. 18 shows the Q value of the output signal light with respect to the fiber length L in the case where a Gaussian filter having a 3 dB bandwidth of 0.2 THz is used as a bandpass optical filter for cutting out the signal light using pump lights having different shapes. Show.

In FIG. 18, the shape of the pump light is as follows: (a) beat light with a repetition frequency of f _B = 160 GHz, (b) Gaussian antiphase with a repetition frequency of 160 GHz and a power half-value width Δt _FWHM of 2.5 ps. A pulse train, and (c) a Gaussian antiphase pulse train having a repetition frequency of 160 GHz and a power half width Δt _FWHM of 1.5 ps are used. The conditions other than the shape of the pump light and the fiber length are the same as those obtained when the results shown in FIG. 15 are obtained. From the results shown in FIG. 18, the maximum value of Q is larger in the order of (c), (b), (a). From this result alone, the shape of the pump light is pulse type and the time width is smaller. However, although the waveform reproduction ability seems to be high, the fact is different from that as described below.

  FIG. 19 shows intensity time waveform eye patterns of signal light output at the fiber length giving the maximum Q value in FIG. 18 for each of the above three types of pump light shapes. FIGS. 19A, 19B, and 19C correspond to FIGS. 18A, 18B, and 18C, respectively. When the waveform quality is evaluated from the eye pattern, power fluctuation at time t = 0 ps is quantified as the Q value. In FIG. 19, regarding the power fluctuation at t = 0 ps, the result of (c) is the best, and this fact is reflected as a large Q value shown in FIG. However, when power fluctuations at times other than t = 0 ps are taken into consideration, the result when the beat light of (a) is used is the best. Considering the total signal quality, it is desirable in the order of (a), (b), (c). The reason why the intensity fluctuation at times other than t = 0 ps becomes large when pulse-shaped pump light having a small time width is used will be described below.

  FIG. 20 shows intensity time waveforms when signal light is cut out by OBPF, which is a Gaussian filter having a 3 dB bandwidth of 0.2 THz, after the parametric process for each of the above three types of pump shapes. 20A, 20B, and 20C correspond to FIGS. 18A, 18B, and 18C, respectively. The solid line shows the signal light waveform when the fiber length L is L = 0.5 km, and the dotted line shows the signal light waveform when the fiber is incident (L = 0 km). Therefore, the waveform at the time of fiber incidence indicated by a dotted line is the same in all cases (a), (b), and (c).

In FIG. 20, when attention is paid to the pulses whose specified times are t ₀ = 0 ps and 6.25 ps, when the dotted line and the solid line are compared, the waveforms deteriorate in the order of (a), (b), and (c). I understand that. In particular, in the case of (c), the component in the vicinity of t = 3 ps that is the valley of the specified time is large, and it is no longer possible to discriminate between two pulses having the specified times t ₀ = 0 ps and 6.25 ps. . In FIG. 19C, the components other than t = 0 ps, that is, the components of the valleys between pulses are increased due to such a phenomenon.

In order to investigate this phenomenon, in the case shown in FIG. 20, the result when OBPF is replaced with an ideal filter having a pass width of 2 THz is shown in FIG. As in FIG. 20, the solid line indicates the signal light waveform when the fiber length L is L = 0.5 km, and the dotted line indicates the signal light waveform when the fiber is incident (L = 0 km). For each waveform when L = 0.5 km shown by the solid line in FIG. 21, when attention is paid to the pulses whose specified times are t ₀ = 0 ps and 6.25 ps, the waveform distortion is (a), (b). , (C) in order.

  Only the pulses with the specified times of 0 ps and 6.25 ps in FIG. 21C are enlarged and shown in FIG. In FIG. 22, the intensity time waveforms of the signal light at L = 0.5 km, the signal light at the time of incidence, and the pump light at the time of incidence are indicated by a solid line, a dotted line, and a broken line, respectively. Under conditions until the parametric process reaches saturation, the gain increases as the instantaneous intensity of the pump light increases, and conversely, the gain decreases as the pump light intensity decreases. That is, the gain is greatest at a specified time when the intensity of the pump light has a peak value, and the gain decreases as the distance from the specified time increases. When pump light with a short time width is used, the temporal change in pump light intensity is steep, and the parametric gain decreases rapidly as it deviates from the specified time. From FIG. 22, when the signal light has a large timing fluctuation at the time of incidence, only a part of the signal light component that is close to the specified time is gained and amplified, while the component that is away from the specified time receives the gain. As a result, the waveform of the signal light is distorted.

  On the other hand, as in FIG. 22, only the part of the pulse whose specified times are 0 ps and 6.25 ps in FIG. 21A is enlarged, and the pump light waveform at the time of incidence is shown by a broken line in FIG. Show. In FIG. 23, the pump light is sinusoidal beat light, and the intensity gradually decreases with the peak at a specified time. As a result, even if the incident signal light has a large time shift, a parametric gain is obtained as the entire signal light component, and no large waveform distortion is observed. Such an effect is obtained as the time width of the pump light is increased, and the effect is maximum when sinusoidal beat light is used.

  As shown in FIGS. 21 to 23, it was found that the waveform of the signal light is distorted when the time width of the pump light is short and the signal light has a large time fluctuation. Returning to the results shown in FIG. 20, the component of the signal light distorted when the time width of the pump light is short passes through the narrow band OBPF and is band-limited, so that it is converted into a component between valleys of pulses. You can see that More specifically, when the time width of the pump light is short, the degree of temporal change in the pump light intensity is steep, and therefore, the phase shift due to XPM generated between the signal having a large time fluctuation and the pump light. Time change also becomes steep. As a result, the instantaneous frequency of the component at the peak time position of the signal light changes more greatly, the optical spectrum spreads, and the intensity is greatly suppressed when passing through a narrow band OBPF. On the other hand, the pulse valley component is not affected by XPM because the intensity of the pump light is almost 0, and passes through the OBPF as it is. Therefore, the peak component at the peak time position is suppressed while keeping the valley component between the signal light pulses, and as a result, the component at the valley is relatively larger than the component at the peak time position.

  As a result of the above process, when the time width of the pump light is short, a large fluctuation component is generated between the valleys of the pulses as shown in FIG. 19C. After all, when applying the waveform reproduction method of the present invention under realistic conditions, it can be concluded that it is best to use sinusoidal beat light as pump light, considering the overall waveform quality. .

  Further, when considering the quality of the signal light input to the waveform regenerator of the present invention, the greater the time width of the pump light, the greater the allowable amount of timing fluctuation, and the maximum when beat light is used as the pump light. . As a result, when performing waveform reproduction relay transmission by applying the waveform reproduction method of the present invention, it is possible to increase the repeat interval by using beat light rather than using pulse-shaped pump light with a short time width. It is economical in considering the cost of the transmission system. However, the pump light used in the present invention is not limited to beat light. That is, in the cases shown in 21 to 23, the signal light has a relatively large time fluctuation. For example, when the timing fluctuation of the signal light is small, even when a pulse having a small pulse width is used as the pump light, It is considered that the same result as that obtained when beat light is used as pump light can be obtained. Therefore, the time width of the pump light may be appropriately selected in consideration of the assumed timing fluctuation of the signal light.

  The above is the elucidation of the phenomenon related to the optical waveform reproduction method of the present invention and the explanation of the optimum design method. Based on the knowledge obtained here, the following is a numerical value for signal regeneration in which the present invention is applied to an ultrahigh-speed optical signal whose bit rate is 160 Gbit / s and quality is degraded by long-distance optical fiber transmission. The effect is confirmed by simulation and experiment.

First, the result of numerical simulation is shown. A signal to be transmitted has a bit rate of 160 Gbit / s, each optical pulse has a Gaussian shape with a power half width of 2.1 ps, and a pseudo random code (Pseudo random bit stream) using an RZ-OOK or RZ-DPSK modulation method. ; Data modulated by PRBS). The group velocity dispersion value at a wavelength of 1.55 μm is 4.67 ps / nm / km, the dispersion slope is 0.05 ps / nm ² / km, the propagation loss is 0.2 dB / km, and the nonlinear constant is 1.46 W ⁻¹ km ^−1. A non-zero dispersion shifted optical fiber 68 km, a group velocity dispersion value of −90 ps / nm / km, a dispersion slope of −0.965 ps / nm ² / km, a propagation loss of 0.2 dB / km, and a nonlinear constant of 2. One span of an optical transmission line is formed by a dispersion-compensating optical fiber of 03W ⁻¹ km ⁻¹ and a length of 3.52 km, and an optical transmission line consisting of a total of three spans is considered. In the optical transmission line, optical amplification is applied for each span, and the average optical signal input power to each span is 10 dBm.

  FIG. 24 shows (a) an optical signal intensity waveform after three span transmission when RZ-OOK is used as a modulation method, and (b) an intensity waveform eye pattern of an electric signal obtained by receiving light with a photodiode. Show. Similarly, FIG. 25 shows an eye pattern when RZ-DPSK is used as the modulation method. It is assumed that the electrical signal was obtained by direct detection with a photodiode having a bandwidth 0.7 times the bit rate, and the RZ-DPSK signal was received in balance after passing through a 1-bit delay interferometer. Shall. FIG. 24 shows that the RZ-OOK signal has both timing jitter and amplitude jitter due to long-distance transmission. These jitters are mainly due to the nonlinear effect of the optical fiber constituting the optical transmission line. In particular, the timing jitter is considered to be due to a phenomenon called intra-channel XPM, and the amplitude jitter is called intra-channel FWM. The Q value obtained from FIG. 24B is 7.0. On the other hand, in the result of the RZ-DPSK signal shown in FIG. 25, the timing jitter is small as compared with the result of RZ-OOK, but a large amplitude jitter occurs. This is because, in the case of the RZ-DPSK signal, optical pulses are present in all bit slots, and the intra-channel XPM that causes timing jitter is suppressed, while the intra-channel FWM that causes amplitude jitter increases. It is thought to be because. The Q value obtained from FIG. 25 (b) is 9.55.

As described above, the quality of an ultrahigh-speed signal having a bit rate of 160 Gbit / s deteriorates mainly due to the nonlinear effect of the optical fiber in long-distance transmission exceeding 200 km. The optical reproduction method according to the present embodiment based on the parametric process is applied to the signal with degraded quality to improve the quality. As for the HNLF for generating the parametric process, the nonlinear constant γ is set to 12 W ⁻¹ km ⁻¹ and the length is set to 0.5 km. The pump light is synchronous beat light with a repetition frequency of 160 GHz, and the average power when input to the HNLF is 21 dBm. At this time, the peak power of the beat light is 250 mW. The average power when the signal light is input to the HNLF is 12.5 dBm when the modulation method is RZ-OOK, and 15.4 dBm when the modulation method is RZ-DPSK. These values are the power of the signal light pulse at the specified time. The average value is set to 77.0 mW. By setting the HNLF, the pump light input power, and the signal light input power to the above conditions, the amplitude stabilization mechanism based on the parametric process is optimized as described above. As shown in the results of FIG. 13, if the timing jitter of the signal light is large, the fiber length for optimizing the amplitude stabilization should be set longer than 0.5 km. Since the magnitude of the optical timing jitter is relatively small, the fiber length was kept at 0.5 km. In addition, a Gaussian filter having a 3 dB bandwidth of 0.3 THz is used as the OBPF that cuts out signal light after propagating through the HNLF.

  FIG. 26 shows intensity time waveform eye patterns of (a) the optical signal and (b) the electric signal of the reproduced RZ-OOK signal corresponding to FIG. When attention is paid to the fluctuation of the component in the specified time of the optical pulse representing the “mark”, it can be seen that both the amplitude jitter and the timing jitter are greatly suppressed. The Q value obtained from FIG. 26B is 12.5, which is significantly improved from 7.0 before reproduction. On the other hand, focusing on the “space” portion, which is significantly smaller than the peak power of the pulse in the “mark” portion, and comparing the results of FIG. 26 and FIG. It can be seen that the intensity ratio of the portion increases in FIG. This is the result of growth of spontaneous emission light noise generated during optical amplification and a component called “ghost pulse” generated by intra-channel FWM, gained by a parametric process with synchronous beat light. Such an increased “space” component can be suppressed by applying an “optical thresholder” known by the method of Non-Patent Document 1. However, in the result of FIG. 26, even if such a threshold is not applied, the result that the Q value is greatly improved as compared with before reproduction is obtained by greatly suppressing the fluctuation of the “mark” portion. It is.

  On the other hand, FIG. 27 shows intensity time waveform eye patterns of (a) optical signal and (b) electric signal of the regenerated RZ-DPSK signal corresponding to FIG. In FIG. 25, it can be seen that the amplitude jitter, which was dominant as a signal quality degradation factor, is greatly suppressed. The Q value obtained from FIG. 27B is 28.4, which is a significant improvement from 9.55 before reproduction. When the RZ-DPSK signal has a large timing jitter, when obtaining the parametric gain by the sync beat light, a phase shift by XPM is also obtained at the same time, and the magnitude depends on the time deviation. That is, the time lag is converted into a phase lag, and furthermore, in the PSK signal, the phase lag is converted into an intensity lag at the time of detection, leading to deterioration of the Q value or an increase in the bit error rate. However, in FIG. 27B, the intensity fluctuation of the electric signal is suppressed by greatly reducing the intensity fluctuation of the optical waveform, rather than the effect of increasing the intensity fluctuation of the electric signal by increasing the phase fluctuation of the optical signal. As a result, the Q value is greatly improved.

  When transmitting an RZ-DPSK signal over a long distance, if the optical waveform has a large amplitude jitter as shown in FIG. 25 (a), the amplitude jitter called the Gordon-Mollenauer effect is caused by a nonlinear effect in the transmission path. Due to the phenomenon of being converted into phase jitter by self-phase modulation, the quality after light reception may be further deteriorated through subsequent transmission. Therefore, when the optical waveform reproduction method of this embodiment is applied, even if the timing jitter of the signal light is converted by the XPM with the synchronized beat light and the phase jitter is slightly increased, the amplitude fluctuation is greatly suppressed, It is expected that the Gordon-Molenauer effect that can occur in the optical transmission of the optical signal is reduced, the phase jitter of the signal light is finally reduced, and the signal quality can be improved.

  From the above results, the effect of the present embodiment was verified using numerical simulation.

  Next, an optical waveform regenerator as an example of the present invention was manufactured, and the effect of the optical 3R waveform reproduction was confirmed by experiments. In the following, signal light is generated using a signal light generator, and optical 3R waveform reproduction is performed on the generated signal light using the optical waveform regenerator according to the embodiment. FIG. 28 is a diagram illustrating the configuration of a signal light generator 500 that generates an OTDM signal of 160 Gbit / s. FIG. 29 is a diagram illustrating a configuration of an optical waveform regenerator 200 according to an embodiment that performs optical 3R waveform reproduction on the generated OTDM signal.

  As shown in FIG. 28, the signal light generator 500 includes a data signal of 10 Gbit / s (exactly 9.95328 Gbit / s) from the signal generator 501 and a clock with a repetition frequency of 10 GHz (exactly 9.95328 GHz). Generate a signal. The optical pulse light source 502 is driven by the generated clock signal to generate an optical pulse train having a center wavelength of 1545 nm, a pulse width of 2.0 [ps], and a repetition frequency of 10 GHz.

The generated optical pulse train and data signal are input to the lithium niobate intensity modulator 503. By driving the intensity modulator 503 with a data signal that is a pseudo random code (PRBS) having a pattern length of 2 ³¹ −1 synchronized with the input timing of the optical pulse train, 10 Gbits having a pulse width of 2.0 [ps] are obtained. / S signal light is generated. This signal light was time-division multiplexed by the OTDM multiplexer 504 to finally generate an OTDM signal of 160 Gbit / s. The bit rate employed in this experiment is precisely 159.25 Gbit / s, but is described below as 160 Gbit / s as usual.

  In order to intentionally degrade the quality of this OTDM signal, a 160 Gbit / s OTDM signal is converted into an optical amplifier 505 and a non-zero dispersion shifted optical fiber (NZ) disposed downstream of the optical amplifier 505. -DSF) 506 and a dispersion compensating optical fiber (DCF) 507 are input to a signal quality deterioration unit 510. The signal quality degradation unit 510 is connected in a total of three stages and transmits an OTDM signal.

  The NZ-DSF 506 that transmits an OTDM signal has a dispersion value of about ¼ in the wavelength 1550 nm band as compared with a standard single mode optical fiber (SMF), so that the pulse waveform is less distorted. Therefore, the fiber is more susceptible to quality degradation due to the nonlinear optical effect.

  In order to confirm the effect of the signal quality degradation when the signal quality degradation unit 510 of three stages is transmitted, the eye pattern of the OTDM signal was measured with an optical sampling oscilloscope having a bandwidth of 500 GHz or more. The measurement results are shown in FIG. 30A and 30B show signal waveforms and eye patterns before transmission of the signal quality degradation unit 510, and FIGS. 30C and 30D show signal waveforms and eyes after transmission of the signal quality degradation unit 510. The pattern is shown.

From the results shown in FIG. 30, it can be confirmed that both the amplitude jitter and the timing jitter are increased after the transmission of the three-stage signal quality degradation unit 510, and the signal quality is degraded. As a result of analyzing the jitter amount based on the eye patterns of FIGS. 30B and 30D, the amplitude jitter (σ _P / μ _P ⁾ and timing jitter [ps] of the signal before transmission are 0.083, 0. The amplitude jitter and timing jitter of the signal after transmission through the signal quality degradation unit 510 were 0.124 and 0.414 [ps], respectively. Signal quality degradation was also confirmed quantitatively.

  Next, a 160 Gbit / s OTDM signal having a center wavelength of 1545 nm generated by the signal light generator 500 shown in FIG. 28 is input as signal light to the optical waveform regenerator 200 that performs optical 3R waveform reproduction shown in FIG. After being amplified by the EDFA 270 which is an optical amplifier, the optical coupler 201 is used to branch the light into two branches.

  One of the branched signals passes through a variable optical attenuator (VOA) 260 and a polarization controller (PC) 280 in order, and then reaches the second optical coupler 203. The other branched signal is input to the 160 GHz synchronous beat generator (OPLL) 202 after being delayed in time by an optical delay line 230. The synchronous beat generator 202 is a light source having a configuration described in Patent Document 1, and generates 160 GHz repetitive beat light having a center wavelength of 1559.12 nm synchronized with an input 160 Gbit / s signal. The synchronized beat light passes through the polarization controller 240 and the band-pass optical filter 250 and reaches the second optical coupler 203 which is an optical signal mixing unit.

  The synchronized beat light and the signal light that have reached the second optical coupler 203 are mixed by the optical coupler 203. Here, the power of the mixed synchronous beat light is constant at 80 mW, while the power of the signal light is adjusted by the VOA 260 in the range of 0 mW to 160 mW. The two mixed signals are amplified by the EDFA 220 and then input to the highly nonlinear optical fiber (HNLF) 204a of the optical 3R regeneration unit 210.

  Note that the phase difference of the intensity waveform between the synchronized beat light and the signal light mixed in the optical coupler 203 does not fluctuate and a certain difference must be maintained. Therefore, in the optical path between the optical coupler 201 and the optical coupler 203, the path length is made as short as possible, and insertion of an optical amplifier that causes a path length variation that causes a phase shift is avoided. However, the insertion of the optical amplifier in the optical path inside the synchronous beat generator 202 is not limited to this. This is because the sync beat generator 202 is a device that always generates beat light that is synchronized with the input signal light, even if the optical path length in the sync beat generator 202 varies due to an optical amplifier or the like.

  After the signal is processed due to the nonlinear effect in the HNLF 204a, the signal light having the same wavelength as the input wavelength is C-band (1530 nm-1560 nm) by a band-pass optical filter 205a such as a wavelength division multiple xing (WDM) coupler. The idler signal light whose wavelength has been converted is separated into L-band (1570 nm-1600 nm) ports and output through band-pass optical filters 205b and 205c having a bandwidth of 4 nm.

  In this experiment, first, the optical waveform regeneration effect of the optical waveform regenerator 200 is confirmed by changing the optical delay amount of the optical delay device 230 in the preceding stage of the 160 GHz synchronous beat generator 202. The effect of optical 3R waveform reproduction based on parametric amplification is to confirm that the effect cannot be obtained unless the phase of the signal light and the phase of the sync beat light are matched within a certain range.

  As an experimental condition at this time, the VOA 260 is adjusted so that the signal light power at the input of the second optical coupler 203 is set to 40 mW, which is half of the synchronous beat light power, and the output power of the subsequent EDFA 220 is 27 dBm (about 500 mW). ). Furthermore, the non-linear constant γ of the HNLF 204a is 10 [1 / W / km] and the length is 100 m, 200 m, and 200 m, and the total length is 500 m.

  Then, the waveform of the signal light output from the band-pass optical filter 205b is changed while changing the delay amount (Δt) of the optical delay device 230 arranged in the previous stage of the synchronous beat generator 202 in order to adjust the phase of the synchronous beat light. The result of measurement with an optical sampling oscilloscope is shown in FIG. From the results shown in FIG. 31, it can be seen that the phase difference between the signal light and the sync beat light changes according to the value of the delay amount (Δt), so that the waveform of the signal light changes according to the phase difference.

  In the following experiment, in order to make the phase of the signal light coincide with the phase of the synchronized beat light, the peak value of the waveform of the signal light output from the bandpass optical filter 205b is maximized and the waveform symmetry is improved. The delay amount of the optical delay device 230 was adjusted. The polarization controller 240 arranged at the subsequent stage of the synchronous beat generator 202 was adjusted so that the intensity of the output signal waveform was maximized.

  Next, FIG. 32 shows the waveform of the idler signal light output from the band-pass optical filter 205c, measured using a 50 GHz band photodiode and a 70 GHz band electrical sampling oscilloscope simultaneously with the output signal waveform measurement shown in FIG. . As in the case shown in FIG. 31, the waveform changes according to the delay amount (Δt), and when the peak value of the output signal waveform in FIG. 31 is high and the symmetry is high (for example, when Δt = 0.0 ps), It can be seen that the interval of the waveform of the idler signal light is equal and the signal-to-noise ratio is high. As shown in FIG. 8, the optimum situation for reproducing signal light is not the optimum situation for idler signal light. However, the waveform of the idler signal light can be a good index for adjusting the phase difference between the signal light and the synchronized beat light.

  Considering the principle of optical 3R waveform reproduction based on parametric amplification, in order to maximize the power of the signal light reproduced with the optical 3R waveform, the polarization of the signal light input to the HNLF 204a and the synchronous beat light are identical. It is desirable to do it. Here, the polarization of the signal light and the synchronous beat light is measured before the optical coupler 203 so that the amplitude of the waveform is maximized while measuring the waveform of the signal light output from the bandpass optical filter 205b with an optical sampling oscilloscope. The polarization controller 280 or the polarization controller 240 inserted in the optical path of the signal light or the synchronized beat light is adjusted. Through this adjustment, the signal light input to the HNLF 204a and the polarization of the synchronized beat light are matched.

  When the waveform of the idler signal light output from the bandpass optical filter 205c is used as a reproduction signal, the power of the idler signal light output from the bandpass optical filter 205c or the amplitude of the waveform is maximized. The polarization controller 280 or the polarization controller 240 inserted in the optical path of signal light or synchronous beat light is adjusted before the coupler 203. Through this adjustment, the polarization of the signal light input to the HNLF 204a and the synchronized beat light are matched.

  FIG. 33 (a) shows the optical spectrum at the input of the HNLF 204a when the power ratio of the signal light and the sync beat light at the input of the HNLF 204a is 1: 2, and the output of the EDFA 220 is 26 dBm (about 400 mW). . FIGS. 33B to 33F show the output of the HNLF 204a when the output of the EDFA 220 is changed without changing the power ratio of the signal light and the sync beat light at the input of the HNLF 204a to 1: 2. The output light spectrum is shown. FIG. 33 (b) shows the result when the output of the EDFA 220 is 20 dBm (about 100 mW), FIG. 33 (c) shows the result when the output of the EDFA 220 is 22 dBm (about 160 mW), and the output of the EDFA 220 is 24 dBm. FIG. 33 (d) shows the result when the output is about 250 mW), FIG. 33 (e) shows the result when the output of the EDFA 220 is 26 dBm (about 400 mW), and the output is about 28 dBm (about 630 mW). The results are shown in FIG. 33 (f).

  From the results shown in FIG. 33, it can be confirmed that with the increase in the output of the EDFA 220, in addition to the increase in the power of the primary idler signal light, the generation of the higher order idler signal light and the increase in power. It can also be confirmed that the power of the signal light is saturated. Also from this fact, it is considered that the increase in higher-order idler signal light contributes to the Reshaping effect.

  The effect of the optical 3R waveform reproduction based on the parametric amplification shows that there is an optimum value with respect to the ratio of the signal light power and the synchronous beat light power from FIG. 8 which is the analysis result using the numerical simulation. Therefore, the optimum ratio was searched experimentally.

  In the following experiment, the HNLF 204a uses a nonlinear constant γ of 10 [1 / W / km], and is configured by connecting two fibers having a length of 100 m and 200 m. The procedure for searching for the optimum value is as follows.

While maintaining the output power of the EDFA 220 installed at the subsequent stage of the optical coupler 203 that mixes the signal light and the sync beat light at 25, 26, 27, or 28 dBm, the mixing ratio of the signal light and the sync beat light is changed using the VOA 260. The eye pattern of the signal light output from the bandpass optical filter 205b was measured with an optical sampling oscilloscope. From the measured eye pattern, the above-described output power, amplitude jitter (σ _P / μ _P ) and timing jitter [ps] at the mixing ratio, and peak power of signal light are evaluated.

FIG. 34 shows the evaluation result of the timing jitter σ _T [ps] indicated by the vertical axis with respect to the mixing ratio indicated by the horizontal axis for each output power, and the amplitude indicated by the left vertical axis for the mixing ratio indicated by the horizontal axis for each output power. FIG. 35 shows the evaluation results of the jitter (σ _P / μ _P ) and the peak power of the signal light indicated by the right vertical axis. In each of FIGS. 34 and 35, (a), (b), (c), and (d) are cases where the output power of the EDFA 220 is 25, 26, 27, and 28 dBm, respectively.

  Note that. The results in FIG. 8 are the results when the average power of the synchronized beat light is constant, whereas FIGS. 34 and 35 are the results when the output of the EDFA 220 is constant. Power varies with ratio. Therefore, it should be noted that FIGS. 34 and 35 cannot be simply compared with FIG. Furthermore, the experimental results and the numerical simulation results do not coincide quantitatively because the numerical simulation does not take into account effects such as the dispersion value of the HNLF 204a and the polarization-dependent dispersion amount (PMD).

  From the results shown in FIG. 34, regarding the timing jitter, when the output power of the EDFA 220 is 26 dBm, the output power of all the EDFAs 220 and the signal mixture ratio are all except for the region where the power ratio of the signal light and the sync beat light is small. The input signal light timing jitter is lower than 0.414 [ps].

  Further, from the results shown in FIG. 35, it was confirmed that the amplitude jitter has a mixing ratio between the output power of the EDFA 220 and the signal that is lower than the amplitude jitter 0.124 of the input signal light described above.

  Next, when the output power of the EDFA 220 is 28 dBm and the average power mixing ratio of the signal light and the sync beat light is 1: 1, the peak power has the maximum value (that is, the mixing ratio in FIG. The signal light waveform and eye pattern at point 1.0 are shown in FIG. 36, and the optical spectrum is shown in FIG.

  36A shows the waveform of the input signal light to the EDFA 270, and FIG. 36B shows the waveform of the output signal light from the bandpass optical filter 205b. FIG. 36C shows the eye pattern of the input signal light, and FIG. 36D shows the eye pattern of the output signal light. FIG. 37A shows the optical spectrum of the input signal light to the HNLF 204a, and FIG. 37B shows the optical spectrum of the output signal light from the HNLF 204a.

From the analysis result of FIG. 36 (d), it was found that the amplitude jitter (σ _P / μ _P ) and timing jitter [ps] of the output signal light were 0.0835 and 0.255 [ps], respectively. While the amplitude jitter and timing jitter of the input signal light are 0.124 and 0,414 [ps], respectively, it can be confirmed that the amplitude jitter and timing jitter are reduced by about 30% and 38%, respectively. . On the other hand, from the result shown in FIG. 37, it can be confirmed that the power of higher-order idler signal light is increased as compared with the optical spectrum shown in FIG.

  Further, FIG. 38 shows a signal light waveform and an eye pattern when the output power of the EDFA 220 is 27 dBm and the average power mixing ratio of the signal light and the sync beat light is 1: 1. Note that FIGS. 38 (a), (b), (c), and (d) correspond to FIGS. 36 (a), (b), (c), and (d), respectively.

From the result of FIG. 38D, the amplitude jitter (σ _P / μ _P ) and the timing jitter [ps] are 0.103 and 0.325 [ps], respectively. Therefore, it can be confirmed that amplitude jitter and timing jitter are reduced by 16% and 21%, respectively, with respect to the input signal light.

  As described above, the optical waveform regenerator according to the present embodiment can confirm the effect of the optical 3R waveform reproduction based on the parametric amplification between the synchronized beat light and the signal light not only in the numerical simulation but also experimentally. It was.

(When cutting out idler light)
In previous studies performed using numerical calculations, after the parametric process, the pump light and idler light are removed by a band-pass optical filter, and only the signal light is cut out, and this is used as reproduced information. Here, a result in the case where the primary idler light is cut out instead of the signal light by OBPF and used as reproduced information is shown.

  Hereinafter, the conditions of the optical fiber for generating the pump light and the parametric process and the conditions other than the peak power of the input signal light are the same as those obtained when the result shown in FIG. 10 is obtained.

Under the above conditions, when the fiber length for generating the parametric process is L = 0.5 [km], the amplitude stabilization operation of the output idler light is optimized, and the pump light and the signal light are optimized. The peak amplitude ratio α ₀ is 0.736 (exact value is 0.73589) as described above. Now, since the peak power of the pump light is set to 250 mW, the peak power of the input signal light may be set to 135.4 mW.

Next, as in the case of FIG. 10, the peak power of the input signal light is changed to any ratio of 70, 80, 90, 100, 110, 120, and 130% with respect to the specified value of 135.4 mW. Then, after combining with the pump light and propagating through the optical fiber of length L, the peak power P when the idler light generated in the vicinity of the wavelength 1566.2 nm is cut out with an ideal filter having a pass band of 2 THz, and these The value of the standard deviation σ _P is calculated from the equation (8) and shown in FIG. In FIG. 39, as in the case of FIG. 10, regarding the idler light peak power when the fiber length is L, the value of the standard deviation σ _P of all trial results is also shown on the right vertical axis.

In FIG. 39, the peak power of idler light is 0 at the initial value of L = 0 and increases with the fiber length L, but the ratio varies depending on the ratio of the input signal light peak power to the specified value. However, at the position where the fiber length is L = 0.5 km, the peak power of idler light converges to a substantially constant value. In fact, the value of the standard deviation σ _P of the idler peak power that was 0 when L = 0 increased with L, then turned to decrease when L> 0.3 km, and decreased to the minimum value σ when L = 0.52 km. _P = 0.83 mW. Shown in FIG. 10, when sigma is compared with the minimum value 1.66mW of _P in the case of cutting out a signal light OBPF after parametric process, sigma _P when cut idler light, it has become a half of the value become. From this, it can be said that the performance of stabilizing the amplitude with fluctuation of the input signal is higher when the idler light is cut out than when the signal light is cut out after the parametric process.

  Next, as in the case of obtaining the result of FIG. 11, the initial time difference between the signal light and the pump light is set to 0.5 ps, the shape of the pump light is (a) beat light, and (b) FWHM is 2.5 ps. Fig. 5 shows the peak time position and first moment of idler light that is output by being cut out by OBPF after propagating through a fiber of length L when it is either Gaussian type and (c) Gaussian type with FWHM of 1.5 ps. Reference numeral 40 denotes a solid line and a broken line. When attention is paid to the value of the primary moment shown by the broken line in FIG. 40, the value of (c) is closer to 0 ps than (b) and (b) at any L. That is, when idler light is cut out by OBPF after the parametric process, it can be seen that the smaller the time width of the pump light, the higher the effect of correcting the time misalignment of the signal light.

  Comparing the primary moment of idler light shown in FIG. 40 with the primary moment when signal light is cut out by OBPF shown in FIG. 11, the idler light is 0 ps more than the signal light. It can be seen that the effect of correcting the time position deviation is higher when the idler light is extracted than the signal light. In the same comparison, the difference between the peak time position and the first moment at L = 0.5 km is the parametric process in all the cases (a), (b), and (c) corresponding to FIGS. When the idler light is extracted after this, the value is smaller, and it can be said that the distortion of the time waveform is smaller. In fact, FIG. 41 shows the idler intensity time waveform at L = 0.5 km and the signal light waveform at L = 0 km by a solid line and a broken line, respectively. However, it can be seen that the waveform distortion is smaller when idler light is extracted by OBPF.

  In FIG. 41, the idler light intensity when the time t is | t |> 2 ps is almost 0 mW, whereas the signal light shown in FIG. Has a certain level of strength. When the bit rate of the signal is 160 Gbit / s, the time when | t | = 3.125 ps is a valley between adjacent pulses on the time axis. When the idler light is generated, the fact that the intensity becomes 0 in the time of this valley means that the waveform shaping effect is high. On the other hand, when the signal light is cut out, the effect is small. Therefore, the waveform shaping effect is higher when idler light is cut out by OBPF after the parametric process. In particular, when the modulation format of the signal is on-off keying (OOK), it is desirable that the intensity component in this valley time is 0, so it is more preferable to cut out idler light.

Next, in FIG. 13, the peak power standard deviation σ _P and the time axis when the signal light is cut out by four types of BPF after the parametric process by changing the initial time difference Δt between the pump light and the signal light. 42 shows the value for the fiber length L of the root mean square <t _m ² > ^1/2 of the first-order moment of the first moment, and FIG. 42 shows the result when the idler light is similarly cut out by a filter. As in the case of FIG. 13, in FIGS. 42A, 42B, (C), (D), and (E), the initial time difference between the pump light and the signal light is Δt = 0, 0.3, 0, respectively. .5, 0.7, and 0.9 ps are shown, and in each case, four types of line types are three types of Gaussian types having different ideal filters and 3 dB bandwidth Δf ₃ dB as OBPF for extracting signal light. It is a result when a filter is applied. Comparing the results of FIG. 13 and FIG. 42, it is common in that the fiber length L (excluding L = 0) at which σ _P is minimum increases as the initial time difference Δt between the signal light and the pump light increases. On the other hand, in the result shown in FIG. 13, the value of <t _m ² > ^{1/2 at} an arbitrary L greatly depends on the shape of OBPF, whereas in the result shown in FIG. 42, the dependency on the shape of OBPF There is almost no. Further, the value of <t _m ² > ^1/2 at L = 0.5 km in FIG. 42 almost coincides with the result in the case of using the 0.2 THz Gaussian filter having the smallest 3 dB bandwidth in FIG. Yes. In other words, when the signal light is cut out by OBPF after the parametric process, the time position deviation correction effect is better when the band of the OBPF is narrow, but when idler light is cut out, the OBPF having a wide band is used. Even so, it can be said that the effect of correcting the time misalignment remains high.

  From the above results, it has been found that it is better to cut out idler light, rather than cut out signal light with OBPF after the parametric process, in terms of amplitude stabilization, time position deviation correction, and waveform shaping performance.

Next, a numerical simulation under a realistic condition is performed, and a state of waveform reproduction when the idler light is cut out by OBPF after the parametric process is examined. FIG. 18 shows three forms of pump light: (a) beat light, (b) antiphase Gaussian pulse train with a power half width of 2.5 ps, and (c) antiphase Gaussian pulse train with a power half width of 1.5 ps. Considering the type, the signal light was cut out by a Gaussian BPF with a 3 dB bandwidth of 0.2 THz after the parametric process under the condition that random fluctuations were given to the amplitude and time position of each pulse in the input signal optical pulse train. In this case, the calculation result of the Q value with respect to the fiber length L is shown. FIG. 43 shows the result of the same calculation performed when the idler light is extracted. 43, as in FIG. 18, the shape of the pump light is as follows: (a) beat light whose repetition frequency is f _B = 160 GHz, (b) repetition frequency is 160 GHz, and half-power width Δt _FWHM is 2. A Gaussian antiphase pulse train of 5 ps and (c) a Gaussian antiphase pulse train having a repetition frequency of 160 GHz and a power half-value width Δt _FWHM of 1.5 ps are used.

In FIG. 43, regardless of the shape of the pump light, the maximum Q value is obtained in the vicinity of the fiber length L = 0.52 km. Focusing only on the maximum Q value, the case of (b) is the best. On the other hand, FIG. 44 shows the intensity time waveform eye pattern of idler light output when the fiber length L = 0.52 km. From FIG. 44, it seems that the case of (c) is the best when both the amplitude fluctuation and the timing fluctuation are taken into consideration. In order to confirm this, FIG. 45 and FIG. 46 show the peak amplitude standard deviation σ _P and the peak time position standard deviation σ _T of the output idler light with respect to the fiber length L, respectively. Note that (a), (b), and (c) in FIGS. 45 and 46 all correspond to (a), (b), and (c) in FIG. 43. From FIG. 45, it can be said that in the fiber length L = 0.52 km, the magnitude of σ _P , which means the degree of amplitude fluctuation, is substantially equal to (a), (b), (c). On the other hand, from FIG. 46, it can be seen that the magnitude of σ _T which means the degree of time fluctuation is clearly small in the case of (c), and timing fluctuation can be largely suppressed. From these results, the overall signal quality when the idler light is cut out by OBPF after the parametric process is best when the pump light has a Gaussian pulse shape and the time width is the smallest 1.5 ps (c). It is.

  As described above, in the case where the signal light is cut out after the parametric process, the result is that the pulse width of the pump light is preferably large. Therefore, depending on whether the idler light or the signal light is cut out, the direction of the pulse width of the pump light to be set is opposite.

  The reason why this occurs is that the waveform reproduction process for the existing signal light and the growth process of idler light generated from zero intensity are fundamentally different. As for the existing signal light, its waveform is transformed by the parametric process with the pump light and the band limiting effect by XPM and OBPF, so the amplitude and time fluctuation of the input signal light are converted into distortion of the output signal light. The waveform reproduction performance may be degraded. In particular, the degree of timing fluctuation of the input signal light can greatly affect the quality of the output signal light. On the other hand, in the latter case, idler light is generated only in a place where the pump light and the signal light overlap with each other in time, so that it is not easily affected by the temporal fluctuation of the input signal light. Furthermore, timing fluctuation can be suppressed more when the pulse width of the pump light is smaller. As a result, the effect of waveform reproduction is enhanced. In other words, the time width becomes the largest as the shape of the pump light in the case of the beat light. At this time, the pulse width is half of the bit slot, so when the idler light is cut out by OBPF after the parametric process, By using pulse-shaped pump light whose pulse width is less than half of the bit slot, the effect of waveform reproduction is enhanced.

100, 200, 900 Optical waveform regenerator 101, 901 Optical demultiplexer 102 Beat light generation unit 103, 903 Signal mixing unit 104, 904a, 904b Nonlinear medium 105, 205a-205c, 250, 905a, 905b Bandpass optical filter 110 , 210 Optical 3R regeneration unit 201, 203 Optical coupler 202 Synchronous beat generator 204a HNLF
230 Optical delay device 240, 280 Polarization controller 260 VOA
220, 270, 505 EDFA
310, 310a Pump light 320, 320a Signal light 330 Primary idler light 331, 332 High-order idler light 401-403 Waveform 500 Signal light generator 501 Signal generator 502 Optical pulse light source 503 Intensity modulator 504 OTDM multiplexer 506 NZ-DSF
507 DCF
510 Signal Quality Degradation Section 910 Clock Extractor 920 Short Pulse Light Source R Range

Claims (26)

Pump light that has the same repetition frequency as the bit rate of the input signal light and a center wavelength different from the center wavelength of the signal light, generates pump light that has a temporally varying intensity and is synchronized with the signal light Generating means;
Signal mixing means for mixing the pump light and the signal light;
An optical waveform reproducing unit that inputs the signal light and the pump light mixed by the signal mixing unit to a nonlinear medium, and generates a nonlinear phenomenon between the pump light and the signal light to reproduce an optical waveform of the signal light; ,
An optical waveform regenerator.   The optical waveform regenerator according to claim 1, wherein both amplitude fluctuation and timing fluctuation of the signal light are suppressed simultaneously.   The optical waveform regenerator according to claim 1, wherein the nonlinear medium is a highly nonlinear optical fiber.   4. The optical waveform regenerator according to claim 1, wherein the nonlinear phenomenon to be generated is time-dependent parametric amplification and cross-phase modulation.   The optical waveform regenerator according to any one of claims 1 to 4, further comprising bandpass optical filter means for extracting the signal light that has undergone the optical waveform regeneration after the nonlinear phenomenon has occurred. .   6. The optical waveform regenerator according to claim 5, wherein the extracted signal light holds phase information at the time of input.   The amplitude ratio of the input signal light and the pump light is set so that the amplitude fluctuation of the input signal light is corrected and the amplitude of the signal light subjected to the optical waveform reproduction is converged to a constant value. The optical waveform regenerator according to any one of claims 1 to 6.   The optical waveform regenerator according to claim 1, wherein the pump light is beat light.   The optical waveform regenerator according to any one of claims 1 to 4, further comprising band-pass optical filter means for extracting idler light generated by the nonlinear phenomenon.   The amplitude ratio between the input signal light and the pump light is set so that the amplitude fluctuation of the input signal light is corrected and the amplitude of the extracted idler light converges to a constant value. The optical waveform regenerator described in 1.   11. The optical waveform regenerator according to claim 9, wherein the shape of the pump light is a pulse shape having a time width shorter than half of a time slot defined by a bit rate.   The optical waveform according to any one of claims 5 to 11, wherein a band of the band-pass optical filter means is reduced to such an extent that no intersymbol interference occurs in the extracted signal light or idler light. Regenerator.   The optical waveform according to any one of claims 1 to 12, further comprising: an optical amplifying unit that amplifies the signal light and the pump light mixed by the signal mixing unit and then inputs them to the nonlinear medium. Regenerator. Pump light that has the same repetition frequency as the bit rate of the input signal light and a center wavelength different from the center wavelength of the signal light, generates pump light that has a temporally varying intensity and is synchronized with the signal light Generation process,
A signal mixing step of mixing the pump light and the signal light;
An optical waveform reproducing step of causing the signal light mixed in the signal mixing step and the pump light to enter a nonlinear medium and generating a nonlinear phenomenon between the pump light and the signal light to reproduce an optical waveform of the signal light; ,
An optical waveform reproducing method comprising:   15. The optical waveform reproducing method according to claim 14, wherein both amplitude fluctuation and timing fluctuation of the signal light are suppressed simultaneously.   The optical waveform reproducing method according to claim 14, wherein the nonlinear medium is a highly nonlinear optical fiber.   17. The optical waveform reproducing method according to claim 14, wherein the nonlinear phenomenon to be generated is time-dependent parametric amplification and cross-phase modulation.   The optical waveform reproduction method according to any one of claims 14 to 17, wherein after the nonlinear phenomenon is generated, the signal light that has been subjected to the optical waveform reproduction is extracted using a band-pass optical filter means.   19. The optical waveform reproducing method according to claim 18, wherein the extracted signal light holds phase information at the time of input.   The amplitude ratio of the input signal light and the pump light is set so that the amplitude fluctuation of the input signal light is corrected and the amplitude of the signal light subjected to the optical waveform reproduction is converged to a constant value. The optical waveform reproducing method according to any one of claims 14 to 19.   21. The optical waveform reproducing method according to claim 14, wherein the pump light is beat light.   18. The optical waveform reproduction method according to claim 14, wherein idler light generated by the nonlinear phenomenon is extracted using a band-pass optical filter means.   23. The amplitude ratio between the input signal light and the pump light is set so that amplitude fluctuation of the input signal light is corrected and the amplitude of the extracted idler light converges to a constant value. 2. An optical waveform reproducing method according to 1.   24. The optical waveform reproducing method according to claim 22, wherein the shape of the pump light is a pulse shape having a time width shorter than half of a time slot defined by a bit rate.   The optical waveform according to any one of claims 18 to 24, wherein a band of the band-pass optical filter means is reduced to such an extent that no intersymbol interference occurs in the extracted signal light or idler light. Playback method.   The optical waveform reproduction according to any one of claims 14 to 25, further comprising an optical amplification step of amplifying the signal light and pump light mixed in the signal mixing step and then inputting the amplified light to the nonlinear medium. Method.

Images