1. Introduction

Amplified spontaneous emission (ASE) introduced by optical amplifiers and non-linear phase noise continuously deteriorate optical signals propagating in optical fiber links, limiting the maximum transmission distance. Thus, noiseless amplification and regeneration of distorted signals are key functionalities to extend the reach of optical networks. All-optical regeneration has recently shown many advantages in processing coherent-optical signals [1], proving that it can be a viable alternative to costly and power consuming optical-to-electrical-to-optical regenerators.

Until the early 2000s, research on all-optical regeneration was focused only on amplitude regeneration. However, with the advent of phase-shift keying (PSK) modulation formats in commercial systems, it is necessary to regenerate both amplitude and phase of the signal, requiring the development of new techniques. Phase-preserving amplitude regeneration [2] and phase-to-amplitude conversion followed by amplitude regeneration and reconversion to a phase modulated signal [3] are some of the proposed scenarios. A more straightforward option is to regenerate the phase directly by taking advantage of phase squeezing properties of PS amplifiers [4]. Amplitude regeneration is also possible when operating in saturation regime [4,5]. Furthermore, PS amplifiers allow a noise figure below the 3 dB theoretical limit of phase-insensitive amplifiers (PIAs) [6,7], enabling low noise amplification along with amplitude and phase regeneration.

The fundamental operation principle of a PSA-based phase regenerator is the property of amplifying or de-amplifying a signal according to the relation between its phase and the phase of other optical pumps [4]. This behavior leads to different gain values for orthogonal quadrature components in a constellation diagram, and naturally squeezes the phase of the signal towards the axis of maximum gain [8]. Most of the PSA-based phase regenerators reported so far have been based on highly non-linear fibers using non-linear techniques such as four-wave mixing [4,5] and self-phase modulation [5]. Recently, phase regenerators based on cascaded second harmonic and difference frequency generation (cSHG/DFG) in PPLN devices have been attracting considerable interest due to their high non-linear coefficient, compactness, low crosstalk and spontaneous emission, no intrinsic frequency chirp, immunity to stimulated Brillouin scattering and ability to operate at room temperature [8–11]. These devices are produced by periodic reversal of the ferroelectric domains in a lithium niobate chip to enhance the non-linear interaction efficiency through quasi-phase matching (QPM) [11]. Single- and multi-channel phase squeezing [8,12], as well as “black-box” phase regeneration [13,14] of BPSK signals were already demonstrated in PPLN-based PS amplifiers using one non-linear device to generate phase correlated idlers/pumps (copier stage [4,15]) followed by a second one to perform PS operation. Each PPLN device requires precise temperature tuning to ensure that the QPM condition is satisfied at the operating wavelengths [8]. In [13], a PPLN-based phase regenerator for BPSK signals was demonstrated, requiring three different PPLN waveguides. As these devices are sensitive to the polarization, the number of required PPLN waveguides doubles in polarization multiplexed systems. Furthermore, even more devices may be necessary for regeneration of higher order PSK modulation formats such as quadrature PSK [16].

In a previous work [17], we proposed using a single PPLN waveguide with bi-directional propagation to simultaneously generate a phase-correlated idler/pump in one direction (designated as copier direction in section 2, and pump generation direction in section 3), and to perform PSA in the opposite one (PSA direction). As only one PPLN waveguide and a single temperature controller are required, the proposed scheme enables potential savings in terms of costs and energy consumption. In this work, we expand the investigation performed in [17] to include experimental characterization of bi-directional PPLN-based PS amplifiers for three waveguides with different lengths.

2. PSA characterization

In this section the PS properties of a PPLN-based PS amplifier with bi-directional propagation are characterized, evaluating the impact of pump-to-signal power ratio, PSR, total input power in PSA direction (including signal, pump and idler waves, measured in the PS mode), P_PS, and the length of the PPLN waveguide. Due to simpler implementation and prospects for multi-channel operation [12], a degenerate pump configuration for the PS amplifier is adopted. In this configuration, the second harmonic of the pump wave interacts with the signal one through difference frequency generation. The experimental apparatus is depicted in Fig. 1. The light sources used in this experiment consisted in two external cavity tunable lasers (ECTLs) tuned at 1545.5 (pump) and 1548 nm (signal). The two signals were aligned to the same state of polarization, amplified, filtered with a band-pass filter (BPF) with 1 nm bandwidth and combined in a 90% coupler. The optical circulators enabled operation in both propagation directions.

 

Fig. 1 Experimental set-up used to characterize PS amplification. BPF - band-pass filter; EDFA - erbium-doped fiber amplifier; OP – optical processor; OSA – optical spectrum analyzer; PC - polarization controller; VOA - variable optical attenuator.

Download Full Size | PDF

Three different PPLN waveguides with lengths of 3, 4.5 and 6 cm length and doped with 5% of MgO in order to reduce photorefractive damage were used in this experiment. The poling period of all PPLNs was 19 μm. The conversion efficiencies, defined as the SHG output power divided by the square of the coupled input power, were of about 190, 410 and 420%/W, for 3, 4.5 and 6 cm, respectively. The QPM wavelength at 27°C for the 3, 4.5 and 6 cm waveguides was 1545.3, 1545.6 and 1545.7 nm. The total insertion losses were about 3.5 dB. The P_PS values obtained in this work were measured before coupling light from the optical fiber to the PPLN waveguide, so they do not account insertion losses. As the two light waves propagate along the PPLN waveguide in the copier direction, a phase-correlated idler is generated with wavelength (λ) and phase (ϕ) relations given by 1/λ_i = 2/λ_p – 1/λ_s and ϕ_i = 2ϕ_p – ϕ_s, where the indices s, p and i stand for signal, pump and idler, respectively [8]. After idler generation, the three waves entered an optical processor (OP) based on a liquid crystal on silicon used to equalize the power, change the relative phase between the interacting waves and switch between PS and phase-insensitive (PI) mode by blocking the idler. Maximum input power constraints and high losses required a variable optical attenuator (VOA) before the OP and an erbium-doped fiber amplifier (EDFA) after it. Then, the interacting waves were sent back to the PPLN for PS operation and to an optical spectrum analyzer (OSA) for analysis. The power of the pump and signal waves entering the PPLN in the copier direction was 25.8 and 14.9 dBm, respectively. A total power up to 27.6 dBm was launched into the device in the PSA direction.

2.1. Results and discussion

The comparison between experimental results and numerical simulations using the model presented in [10] describing the cSHG/DFG interaction is depicted in Fig. 2. In the numerical simulations, it was assumed identical insertion losses for both sides of the PPLN device.

 

Fig. 2 Experimental results for the devices with lengths of: (a), (d) and (g) 3 cm; (b), (e) and (h) 4.5 cm; (c), (f) and (i) 6 cm. (a)-(c) Experimental and simulated PS gain as a function of the phase added to the pump wave in the OP for P_PS of 27.6 dBm and PSR of 10 dB. (d)-(f) Experimental and simulated PSDR curves as a function of the PSR for P_PS = 27.6 dBm. (g)-(i) Maximum experimental PS gain (asterisks) and attenuation (triangles), simulated (solid curve) and experimental (circles) PSDR and QPM temperature (dots) as a function of P_PS, for PSR = 10 dB.

Download Full Size | PDF

Phase sensitive operation is clearly evident in results shown in Figs. 2(a)-2(c). The PS gain, defined as the ratio between the signal power in PS and in PI modes, depends on the phase added to the pump wave, exhibiting maximum gain and attenuation (minimum gain) peaks for phase values of around 1.6π and π, respectively. The phase sensitive dynamic range (PSDR), defined as the ratio between the maximum gain and maximum attenuation [8] reached values of approximately 5.3, 8.6 and 8.8 dB for the devices with 3, 4.5 and 6 cm, respectively. In Figs. 2(d)-2(f), the experimental results and numerical simulations show that the PSDR increases with the PSR, but for PSR values higher than 15 dB this variation is small. In Figs. 2(g)-2(i), the results indicate that the PSDR increases with P_PS. As higher P_PS means higher pumping power for the same PSR value, this is an expected result. For the 3 cm waveguide a good agreement between experimental data and simulations is verified. For the other devices, the experimental values are up to 3 dB lower than the simulations. This discrepancy is thought to originate from power loss resulting from sum-frequency generation between the SHG and the input signals. In these devices, intense green light emission was observed for high pumping powers as previously. This phenomenon was also observed in [8], where it was suggested that it is the result of sum-frequency generation between the SHG and the input signals. Green light emission along with green light induced infrared absorption (GRIIRA) in lithium niobate depletes the power of the second-harmonic and decreases the PSDR [8,18]. In addition, the QPM temperature changed due to residual photorefractive damage. As higher PSDR means better phase regenerative properties [8], higher PSR and P_PS would be desirable. However, these effects cause a trade-off where, on one hand, increasing pump power results in stronger green light emission, GRIIRA and photorefractive damage. On the other hand, reducing the signal power degrades the optical signal-to-noise ratio.

3. Black-box regenerator for BPSK signals

In this section, a bi-directional PPLN-based “black-box” phase regenerator for BPSK signals is investigated. The set-up used in these experiments is depicted in Fig. 3. In order to enable “black-box” operation, a non-degenerate pump configuration for the PS amplifier is adopted in this section. This configuration allows the generation of a modulation-free phase-correlated idler in the pump generation stage that will later act as the second pump (Pump2) for PSA. Although this configuration differs from the one characterized in section 2, the main qualitative conclusions of the influence of parameters such as P_PS and PSR are still valid. In this case, it is the second harmonic of the input signal that interacts with the pump waves through DFG. It should be noted that the wavelength and phase relations of the generated wave are now given by 1/λ_p2 = 2/λ_s – 1/λ_p1 and ϕ_p2 = 2ϕ_s + 2ϕ_sN – ϕ_p1, where the indices s, p1, p2 and sN stand for signal, Pump1, Pump2 and signal noise, respectively. Therefore, the BPSK modulation is stripped, but an excess phase noise 2ϕ_sN is transferred to the generated second pump [4]. In order to reduce the excess phase noise and improve carrier recovery performance, an injected-locked semiconductor laser is introduced in the set-up as well as an electrical phase-locked loop (PLL) and a PZT fiber stretcher to compensate slow path length deviations [4].

 

Fig. 3 Experimental set-up for the bi-directional PPLN-based black-box regenerator with injection-locking for carrier recovery. MUX/DEMUX – WDM multiplexer/demultiplexer; ODL – optical delay line; OP – optical processor; Rx – coherent receiver.

Download Full Size | PDF

A noisy 10 Gb/s phase modulated signal was generated by modulating the light emitted by an ECTL with a phase modulator (PM) driven by a noisy electrical signal. A pseudo random bit sequence of length 2¹⁵–1 was used in this experiment. Additional white noise was provided by a noise source composed by two cascaded EDFAs interleaved by a 3 nm BPF. The contribution of each noise source was selected in order to obtain standard deviation values of the detected symbols for phase and amplitude of 0.25 and 0.1, respectively, before entering the PPLN for phase regeneration. Part of the modulated signal was combined with the Pump1 wave emitted from another ECTL in a WDM coupler and injected into the PPLN device. In this experiment, a PPLN that was not characterized in the previous section was selected. The new PPLN was chosen as a trade-off between the spectral operational range of the injection-locking laser and the pass band of the WDM coupler, while enabling room temperature operation. The PPLN waveguide length, poling period, QPM wavelength at 21°C, MgO content and SHG conversion efficiency were 6 cm, 19.1 μm, 1549.3 nm, 5% and 560%/W, respectively.

After generation of the Pump2 wave in the PPLN, the excess phase noise was reduced by the semiconductor injection-locked laser as in [4]. The WDM multiplexer/demultiplexer (MUX/DEMUX) was used to filter all the signals except Pump2 at the input to the semiconductor laser and also to combine its output signal with the other pump coming from the Pump1 arm. Then, the phase-correlated pumps were combined with the modulated signal from the Signal arm and sent to an OP and an EDFA for power equalization and amplification before entering the PPLN for phase regeneration. The function of this EDFA is to compensate for the attenuation required at the input of the OP due to maximum input power constraints and for passive losses. As in the previous section, the OP allowed switching from PI to PS mode by blocking the pump2 signal. The input power going into the PPLN waveguide in the pump generation direction was about 27.3 dBm whereas in the PSA direction it was varied up to a maximum value of 28.9 dBm. Optical delay lines (ODLs) and polarization controllers (PCs) were also introduced in the Signal and Pump1 arms to guarantee path length and polarization alignment.

A part of the regenerated signal was tapped from a 10% tap located between the PPLN and the WDM coupler and sent to a variable bandwidth BPF and a single polarization coherent receiver. The coherent receiver was composed by a 100 kHz linewidth ECTL used as the local oscillator and a 90° optical hybrid received with DC-coupled photodiodes at the input to a 40 GSample/s real-time sampling oscilloscope with a 13 GHz bandwidth. The deviations of the equalized path lengths were compensated by a PLL shown in the phase locking control box in Fig. 3 and a PZT fiber stretcher in the Pump1 arm, as described in [6]. The feedback error signal for the PLL was obtained from a fraction of the regenerated signal after the PSA stage.

3.1. Results and discussion

In order to evaluate the performance of the regenerator the ratio between the standard deviations of the received symbols in PI and PS mode (σ_PI/σ_PS) for both amplitude and phase was measured, for different values of PSR and P_PS. The total number of received symbols used to obtain the standard deviation values was 4096. It should be noted that even though the nomenclature for the signal and pump wave is different in this section, PSR is still the pump-to-signal power ratio and P_PS includes the power of the signal, Pump1 and Pump2 waves. In the PSA stage, no significant changes on the standard deviation values for the phase and amplitude of the detected symbols were observed before and after the PPLN device, in PI operation mode. Therefore, the ratio σ_PI/σ_PS is a simple and accurate measurement of the regenerative properties in the PS mode. Effective phase or amplitude regeneration occurs when σ_PI/σ_PS > 1, with better regenerative performances for higher σ_PI/σ_PS. The obtained results are depicted in Fig. 4.

 

Fig. 4 Ratio between the standard deviations of the received symbols in PI and PS operation modes for amplitude and phase as function of: (a) PSR with P_PS = 28.9 dBm; (b) P_PS with PSR = 10 dB. Received constellations (c) and phase histograms (d) of the received symbols in PI and PS modes.

Download Full Size | PDF

For all the measurements performed (Fig. 4(a) and Fig. 4(b)) the ratios σ_PI/σ_PS in terms of phase are higher than 1, proving effective phase regeneration when the operation mode switched from PI to PS. In addition, higher PSR and P_PS values resulted in better phase regenerative performances, which is in agreement with the PSDR measurements shown in the previous section. The phase squeezing characteristic of the bi-directional PPLN-based PS amplifier is evident from the constellation diagrams depicted in Fig. 4(c) and the phase histograms shown in Fig. 4(d). By switching from PI to PS mode, the phase of the received symbols is squeezed towards the real axis of the constellation diagram and the distribution of the angles of the ‘0’ and ‘π’ symbols becomes narrower. Amplitude regeneration would be expected for PSR ratios close to 0 dB as the PS amplifier operates in the saturation regime [15], but it was not observed in this experiment. Even though higher pumping power results in better phase regeneration, the results shown in Fig. 4(b) suggest that additional amplitude noise is also introduced, as a result of phase-to-amplitude noise conversion [15].

The regenerator studied in this section also suffered from green-light induced refractive index change and GRIIRA, affecting both PS gain and system stability. Refractive index variation modifies the optical path length of the Pump1 + Signal arm and the generation of the second pump, affecting both injection locking and the operation of the PLL. Although the time during which the system remained stable was greater than any measurement time, improvements on thermal and refractive index stability of the PPLN must be pursued in order to make bi-directional PPLN-based PSA a viable technique for phase regeneration.

4. Conclusion

An experimental investigation of a PSA set-up and in-line phase regenerator relying on a single bi-directional PPLN for both generation of phase-correlated signals and PS operation is reported. In both cases, best performance in terms of PSDR and phase regeneration occurred for higher PSR and P_PS values.

A trade-off between high PS gain provided by high pump power and system stability was observed due to green-light induced refractive index changes, photorefractive damage and GRIIRA. Moreover, the bi-directional nature of the proposed scheme enhances this limitation. Therefore, it is crucial finding a way to mitigate these issues as well as developing new devices with higher efficiencies to reduce high pump power requirements. One possible solution is to replace the annealed proton-exchange PPLN devices used in this work by direct-bonded ZnO-doped PPLN ridge waveguides, as in [19]. Besides their higher non-linear efficiency, direct-bonded ridge waveguides are more resistant to deleterious photo-induced effects.

The results presented in this work show that PSAs can be implemented in a single PPLN waveguide with bi-directional propagation. Further improvements in the set-up/PPLN waveguide and the possibility of photonic integration will reduce the amplification requirements and allow savings of devices, temperature controllers and complexity. These advantages become especially attractive for regeneration of multi-level phase-encoded signals as BPSK regenerators have been suggested as elementary building blocks [4,16] and in polarization multiplexed systems.