1. Introduction

In a long-haul transmission system with optical amplifiers as repeaters, the dominant factors limiting the achievable transmission distance and channel capacity are the degradation of the signal to noise ratio (SNR) due to the accumulation of optical noise from the repeater amplifiers and inter-symbol interference caused by nonlinear impairments in the transmission optical fibers [1, 2]. To enhance the SNR of such inline optical repeater systems, phase-sensitive amplifiers (PSA) are now attracting a great deal of interest because of their potential for low noise amplification and their signal regeneration capability with a better SNR than a conventional phase-insensitive amplifier (PIA) such as an erbium doped fiber amplifier (EDFA). The first experimental demonstration of low noise amplification with a noise figure below the 3 dB quantum limit using a nonlinear fiber has been reported [3]. Phase regeneration can be achieved by using the phase squeezing capability of a PSA [4]. Amplitude regeneration can also be obtained by operating the amplifier in the saturated regime [5]. These capabilities mean that PSAs will have a large impact on long-haul transmission if they are used as multiple repeaters.

There have already been several studies on transmission using a PSA as a single repeater. Phase and amplitude regeneration has been demonstrated using a fiber-based in-line PSA in an 800 km dispersion managed link with the regenerator placed in-line or at the receiver [6]. The fiber-based PSA requires a long interaction length of over a few hundred meters because of the small χ⁽³⁾ nonlinearity of the optical fiber. In terms of phase locked loop (PLL) design, one of the most important issues is loop delay. If we are to achieve a wide feedback bandwidth with a short loop delay we must minimize the total length of the PLL. A phase difference between the pump and signal is detected at the output of the nonlinear medium utilizing the phase sensitive gain, and so a shorter interaction length is preferable. In addition, the length of the carrier recovery and the pump generation stage of in-line PSA must be short to achieve stable phase locking. This is because phase fluctuations caused by the path difference between the data signal and the carrier recovery/pump generation stage can be minimized by using a shorter interaction length. Therefore, the longer interaction length may lead to difficulties as regards the stabilization of the phase locking between signal and pump. On the other hand, recent advances on periodically poled LiNbO₃ (PPLN) have enabled us to construct a χ⁽²⁾-based PSA using waveguides with a very short length of only a few centimeters. Recently, we have demonstrated the first in-line operation using a PPLN-based PSA with a phase regeneration capability. The in-line PSA operates successfully as a repeater amplifier in a 160 km dispersion managed link [7]. The signal regeneration capability of the PSA will be useful if it is implemented as a repeater amplifier for multi-span transmission. However, multi-span transmission using a PSA has yet to be demonstrated because of the insufficient stability of the PSA. To date, there has been no experimental investigation of ultra long-haul transmission using a PSA as a repeater amplifier.

More recently, we demonstrated multi-span transmission using the PPLN-based in-line PSA for the first time [8]. We achieved transmission over 3200 km without in-line dispersion compensation by regenerating the phase and amplitude of a signal degraded by chromatic dispersion, fiber nonlinearity, and the amplified spontaneous emission (ASE) of optical amplifiers. However, the transmission characteristics were not evaluated under individually optimized conditions for either the PSA link or the EDFA reference link, with respect to fiber input power and residual dispersion. Only a fixed high fiber input power condition was examined because the available fiber input power was limited by the inflexibility of the power level diagram resulting from the long transmission fiber span. In particular, the transmission distance of the EDFA link as a reference for comparison was significantly limited. Therefore, further research on the efficacy of a PSA in multi-span transmission is required if we are to quantify the PSA characteristics.

An important factor for a transmission system is its degree of tolerance to nonlinear distortions. A simple way to achieve a signal with a high SNR is to increase the signal intensity. However, the transmission reach will ultimately be limited by the interplay between the Kerr effect in the fiber and amplified spontaneous emission (ASE) from the amplifiers, such as the Gordon-Mollenauer effect [9].

In this paper, we describe an improvement in the tolerance to nonlinear distortions in a multi-span transmission using the PPLN-based in-line PSA. In our recent implementation of the in-line PSA [10], we realized a fast PLL to cope with burst data using injection locking and electro-optic (EO) modulator based phase dithering. This enabled us to examine the performance of the PSA as a multi-repeater in a recirculating loop transmission experiment. To achieve multiple repeater operation, we have improved the gain of the PPLN-based PSA by modifying the module design to realize a high external gain of 12 dB. Thanks to the phase and amplitude regeneration capabilities of the PSA, the inter-symbol interference caused by the phase noise resulting from fiber nonlinearity and the intensity noise caused by the ASE of an optical amplifier are largely suppressed. This enables us to achieve the 5460-km ultra long-haul multi-span transmission of a 28-Gb/s binary phase shift keying (BPSK) signal. We examined the Q-factor dependence on transmission distance for both a PSA and an EDFA at the same link. The Q-factor degradation for the PSA link was significantly mitigated compared with that for the EDFA link. For the PSA link, a Q-factor advantage of 8.5 dB was obtained after a 4200 km transmission at a 12 dB higher signal power than the optimum power for the EDFA link.

2. Experimental setup and characteristics of in-line PSA

We examined the characteristics of the in-line PSA in a multi-span transmission using a recirculating loop transmission. Before describing the experimental setup, we show the transmission line design concept. In a previous report [8], we used an 80-km dispersion-shifted fiber (DSF) without dispersion compensation. In this experiment, we used a 42-km dispersion compensated transmission line. For the PSA link, the frequency chirp due to the chromatic dispersion can be regenerated [8, 11]. While for the EDFA link, post-compensation using coherent detection and digital signal processing (DSP) is very effective in chromatic dispersion compensation [12]. However, a direct comparison of these effects is difficult when there are large nonlinear impairments. We focused on an examination of the phase and amplitude regeneration effects for a signal degraded by the phase noise resulting from fiber nonlinearity and the intensity noise caused by the ASE, because these noises are dominant factors in SNR degradation. Therefore, we used a dispersion compensated transmission line in this experiment. As regards the fiber length for 1 span, we used a short span length of 42 km to provide the fiber with a wide range of input powers. The upper limit of the input power is determined by the nonlinear impairments caused by the fiber nonlinearity. While the lower limit of the input power is determined by the span loss and the minimum amplifiable signal power of an optical amplifier because the NF of the amplifier is degraded if the signal power is too low. If the fiber length of the span is long, which means the span has a large loss, the available input power will be fixed. To retain a wide input power range for examining the transmission characteristics, we used a short span length of 42 km in this experiment.

Figure 1 shows the experimental setup, which utilizes a recirculating loop transmission line. The optical carrier generated from an external cavity laser diode (ECLD) at a wavelength of 1535.8 nm was modulated by a LiNbO₃ Mach-Zehnder modulator (MZM) with the BPSK format at a data rate of 28 Gbit/s. The pattern length of the tested pseudo-random binary sequence (PRBS) was set at 2¹⁵-1. The transmission power is defined as the signal power launched into a transmission fiber in the recirculating loop. The fiber input power could be varied from −10 to + 10 dBm. The transmission line consisted of an EDFA/PSA hybrid in-line amplifier, a 40 km dispersion-shifted fiber (DSF), and a 2 km single mode fiber (SMF). At the test wavelength of 1535.8 nm, the loss of the 42-km span was 10.0 dB, and the residual chromatic dispersion per span was less than 1 ps/nm. The tandem connection of the EDFA and PSA yielding a gain of 21 dB was used to compensate for the loss of the loop. To allow us to use gain saturation for amplitude regeneration, we set the EDFA as a pre-amplifier and the PSA as a post-amplifier.

 

Fig. 1 Experimental setup for recirculating loop transmission.

Download Full Size | PDF

After transmission, the BPSK signal was detected by a digital coherent receiver. The real and imaginary parts of the transmitted signal were detected by balanced photo detectors and digitized at 50 GS/s by a digital storage oscilloscope. The received data were post-processed off-line. Note that there was no compensation for the chromatic dispersion of the entire transmission line in the digital coherent receiver. The time duration of the recirculating data was about 210 μs, which corresponds to the delay of the fiber. For comparison we also undertook a recirculating loop transmission using an EDFA only repeater.

Figure 2 shows the configuration of the in-line PSA for a BPSK signal [10]. There are three PPLN ridge waveguides: the first for carrier recovery, the second for pump generation, and the third for optical parametric amplification (OPA). We use cascading second harmonic generation (SHG) and difference frequency generation (DFG) in PPLN waveguide 1. By doubling the signal phase, the carrier is recovered from a BPSK signal. The carrier phase is copied to a 1.5-μm-band idler wave by the DFG process between the SH wave and a CW wave as LO1 from an ECLD. The idler wave is injected into a semiconductor slave laser as LO2 for phase locking. LO1 and LO2 are injected into PPLN waveguide 2 to generate a sum-frequency (SF) wave at around 770 nm. This corresponds to the SH wavelength of the signal. The SF pump and data signal are injected into PPLN waveguide 3 for OPA. The in-line PSA was equipped with carrier recovery and a PLL. Carrier phase recovery based on the optical nonlinear effect and optical injection locking is very fast compared with the time duration of the sequential burst data. The high-speed locking of the relative phase between the signal and pump is important in terms of achieving a PSA for the sequential burst data. The optical path length of the loop for the PLL is about 60 m in this setup, then the loop delay associated with optical fiber components is about 200 ns. Therefore, we can use the EO modulator based phase dithering at a frequency of 1 MHz. The band frequency of the error signal fed back to the PZT is sufficiently wide compared with the maximum response bandwidth of the PZT, which is several tens of kHz, even after passing through the loop filter because of the high-speed phase dithering. This enables us to achieve stable and high-speed phase synchronization for the recirculating data whose phases between the sequential burst data were discontinuous.

 

Fig. 2 Configuration of in-line PSA with carrier recovery and phase locking system.

Download Full Size | PDF

Here, we improved the performance of the in-line PSA to achieve a higher gain than that previously reported [10]. We modified the design of the SFG module to minimize the coupling loss between the PPLN waveguide and the SFG output fiber. This enabled us to obtain a pump power as high as 500 mW. The PSA gain will increase exponentially with the pump power, meaning that increasing pump power is effective in improving the PSA gain. We also minimize the coupling loss between the PPLN waveguide and the input/output fiber for the signal in the OPA module to reduce the insertion loss of the signal. The result is that the insertion loss of the OPA module is as low as 4.0 dB. Figure 3(a) shows the spectra of the input and output of the in-line PSA with a 10 Gb/s BPSK signal. We obtained a low total loss of 5.7 dB for the in-line PSA. With the in-phase signal, we realized an internal gain of + 17.7 dB. Thus, we obtained a high external gain of + 12.0 dB. With the quadrature-phase signal, we obtained an internal de-amplification gain of −9.3 dB. Accordingly, this provided a phase sensitive dynamic range (PSDR) as high as 27.0 dB. Thanks to the improvement in the PPLN modules, we succeeded in achieving a higher gain and PSDR than in our previous study [10].

 

Fig. 3 (a) PSA output spectra with in-phase or quadrature-phase BPSK signal. (b) Measured external PSA gain vs. input power to in-line PSA.

Download Full Size | PDF

In addition, the gain saturation can be used for amplitude regeneration. The parametric process can achieve bit by bit amplification because of its fast response time. Figure 3(b) shows external PSA gain as a function of input power to the PSA from −20 to + 20 dBm. The gain saturation occurred at an input power of over 0 dBm.

3. Experimental results of PSA repeated optical transmission in recirculating loop

First, we examined the transmission characteristics for the EDFA loop link as a reference. Figure 4(a) shows the Q-factor as a function of recirculation number (top axis) and distance (bottom axis) with some different fiber input power levels for an EDFA loop. The dashed line indicates the transmission limit line at a Q-factor of 8.3, which corresponds to the FEC limit with a 7% overhead. For the EDFA loop link, the Q-factor decreases linearly with distance and the maximum transmission distance was obtained at a fiber input power of −5 dBm. Figure 4(b) shows constellation diagrams at a fiber input power of −5 dBm for 0, 30, 50, 70 iterations, which corresponds to transmission distances of 0, 1260, 2100, and 2940 km, respectively. The signal quality was degraded by ASE noise and fiber nonlinearity during each iteration. The Q-factor reached the FEC limit after a transmission distance of 2940 km, which corresponds to the 70th iteration.

 

Fig. 4 (a) Results for EDFA configuration for Q-factor vs. transmission distance (bottom axis) and recirculation number (top axis). (b) Constellation diagrams at fiber input power of −5 dBm for 0, 30, 50, 70 iterations, which corresponds to transmission distances of 0, 1260, 2100, and 2940 km, respectively.

Download Full Size | PDF

Then, we adopted the in-line PSA for same loop line simply by inserting the PSA after the EDFA. We set the total gain of the tandem EDFA-PSA link at 21 dB, which is the same as the gain of the EDFA link. Figure 5(a) shows the Q-factor as a function of transmission distance for fiber input powers of + 4, + 7, and + 10 dBm. Even at a high input power of + 4 dBm, a maximum distance of about 3000 km was obtained, which is the same as that of the EDFA link. At a fiber input power of + 4 dBm, the signal power to the PSA has not yet reached the power level for gain saturation. This fiber input power of + 4 dBm corresponds to a PSA input power of about + 3 dBm. In this experimental setup, the 10 dB loss of the transmission fiber is roughly compensated by the EDFA pre-amplifier, thus the signal power to the PSA is of almost same order as the fiber input power.

 

Fig. 5 (a) Results for EDFA-PSA configuration for Q-factor vs. transmission distance (bottom axis) and recirculation number (top axis). (b) Constellation diagrams at fiber input power of + 7 dBm for 0, 70, 100, 130 iterations, which corresponds to transmission distances of 0, 2940, 4200, and 5460 km, respectively.

Download Full Size | PDF

The signal power to the PSA will reach the gain saturation power level by increasing the fiber input power even further. At a fiber input power of + 7 dBm, the degradation of the Q-factor is significantly mitigated. The slope of the Q-factor decrease with distance is much lower than that for the EDFA link. For the PSA link, the Q-factor remains high as the number of iterations increases. Figure 5(b) shows the constellation diagrams at a fiber input power of + 7 dBm for each of the iterations of 0, 70, 100, and 130, which correspond to transmission distances of 0, 2940, 4200, and 5460 km, respectively. A reduction in the phase noise by utilizing the phase squeezing property, and the distinct symbol separation resulting from gain saturation were clearly observed even after transmission over 5000 km. In fact, phase and amplitude regeneration increase the maximum transmission distance compared with the EDFA link. The Q-factor was still higher than the FEC limit at the 130th iteration, which corresponds to 5460 km. The reason for the drop in the Q-factor beyond that point is the failure of the PLL. Thus, a longer distance transmission will be achieved by improving the PLL.

We compared the results of the EDFA and the PSA loop links. Figure 6(a) shows the maximum distance and corresponding iteration as a function of fiber input power. The maximum transmission distance of 2940 km was obtained at a fiber input power of −5 dBm for the EDFA link. At a higher input power, the distance was limited because of the non-linear noise in the fiber. Meanwhile for the PSA link, we can increase the fiber input power because it is highly resistant to nonlinear noise. Even when the input power is 10 dBm, the maximum transmission distance is greater than that for the EDFA link. The maximum transmission distance was obtained at a fiber input power of + 7 dBm for the PSA link.

 

Fig. 6 Comparison of the EDFA and PSA links (a) maximum transmission distance vs. fiber input power (b) Q-factor vs. transmission distance.

Download Full Size | PDF

We also compared the results for the EDFA and the PSA loop links in terms of the Q-factor. Figure 6(b) shows the Q-factor as a function of the transmission distance at fiber input powers of −5 and + 7 dBm for the EDFA and PSA links, respectively. The maximum transmission distance was obtained under these input power conditions. The Q-factor decreased linearly with distance for the EDFA loop link. On the other hand, the degradation of the Q-factor for the PSA link was significantly mitigated compared with that for the EDFA link. The difference between the Q-factors for the EDFA and PSA become clear during each iteration. There was a 12 dB difference between the fiber input powers of the EDFA and PSA links in this comparison. If we can disregard the nonlinear effects in the transmission fiber, the signal quality improves as the signal power increases. However, the transmission reach was ultimately limited by the nonlinear impairments for the EDFA link. Under the above conditions, an 8.5 dB Q-factor advantage was obtained after 100 iterations, which correspond to a 4200 km transmission. These results show the effectiveness of the PSA in improving the tolerance to the nonlinear distortions in long-haul transmissions.

4. Discussion

The demonstration of multi-span transmission with phase and amplitude regeneration reported in this work illustrated the consistent progress made on PSA research with the aim of improving the signal quality for long-haul transmission in future high capacity optical communications. However, the in-line PSA using degenerate parametric amplification reported in this work can amplify only one wavelength signal carrier, one polarization, and one of two quadrature phase components of the signal. The high capacity communication of the future will most probably operate with spectral efficiencies of better than 1 bit/s/Hz, thereby requiring the use of multilevel phase coding and polarization multiplexing. The achievement of high capacity communication of course requires the use of a multi-wavelength carrier signal. A PSA for simultaneous multi-channel amplification has been investigated using non-degenerate parametric amplification [13, 14]. A PSA for quadrature phase shift keying (QPSK) signals has also demonstrated using the non-degenerate process [15, 16]. Further study is required to achieve the amplification of signals with higher spectral efficiencies such as dual-polarized quadrature amplitude modulation (QAM) signals. There are also many challenges to be faced if we are to realize stable operation. Simplifying the configuration of the in-line PSA and long-term stabilization of the PLL are thought to be essential. In this sense, the PPLN waveguide may potentially be integrated with several functions on a chip. A few studies have been undertaken on monolithic integration [17, 18]. The integration of several PPLN waveguides on a chip will play an important role in providing a compact device with stable operation and high functionality.

5. Conclusion

We achieved a χ⁽²⁾ based in-line PSA with a high internal/external gain of about + 18 dB / + 12 dB and a wide PSDR of 27 dB by improving the PPLN devices. We reported the demonstration of multispan transmission using a PPLN-based PSA by implementing the in-line PSA as a repeater amplifier for a 28 Gb/s BPSK signal in a recirculating loop. We achieved transmission over 5400 km by regenerating the phase and amplitude of a signal that was degraded due to fiber nonlinearity and the ASE of optical amplifiers. We showed that a PSA has potential to improve signal quality for long-haul transmission.