1. Introduction

Differential phase-shift-keying (DPSK), in which information is encoded in the phase difference between adjacent bits, is becoming increasingly popular for optical transport systems since it offers a better tolerance to fiber nonlinearities and polarization mode dispersion [1]. At the optical receiver, phase information needs to be converted into intensity information prior to detection. For this task, a widespread solution consists of using a delay-line interferometer (DLI) providing a temporal delay of one bit period between the two arms. In this arrangement, either one (single-ended configuration) or the two output ports of the interferometer can be used for direct or balanced detection, respectively. In the linear transmission regime, balanced detection offers a near 3-dB sensitivity improvement with respect to single-ended demodulation [1]. Despite these benefits, the use of a balanced photodetector, together with the complexity required to stabilize the DLI against environmental fluctuations make the implementation of a low-cost DPSK-receiver a challenging task.

DPSK demodulation using optical pass-band spectral filtering followed by single-ended detection has been reported [2–5]. The higher stability has been achieved at the cost of 3-dB sensitivity [1]. However, it is worth noting that the penalty between the single- and double-ended receptions is reduced, especially when non-linear transmission impairments start to play a significant role [6,7]. On the other hand, pass-band optical filtering makes the system more tolerant to dispersion [4,8,9]. Several pass-band optical filtering configurations have been recently reported, including Gaussian-shaped tunable filters [2,3], specially structured fiber Bragg gratings [4,5], or arrayed waveguide grating demultiplexers [7]. The pass-band filtering demodulation scheme has been demonstrated in binary DPSK [2–5] and differential quadrature phase shift keying [7,8] employing non-return-to-zero (NRZ) format. Band-pass filters emulating the constructive and destructive ports of the DLI and thus allowing for balanced detection have been recently reported in [10] using silicon microring resonators.

The above mentioned filters offer no possibility for reconfiguration and therefore can only be implemented in optical fiber networks working at a fixed rate. Future optical networks, however, might require multi-rate operation and multi-format receivers to accommodate the increasing amounts of data traffic. Most of the network providers that today operate at 10 Gb/s will eventually upgrade to 40 Gb/s and beyond. It is expected that multiple transmission rates will coexist for better resource exploitation [11]. Unfortunately, the flexibility required for this kind of optical network challenges most of the previously mentioned approaches to demodulate DPSK signals. Several attempts to design tunable bit-rate DPSK receivers have been reported, such as a DLI implemented in a birefringence tunable fiber [12], where the differential group delay can be continuously tuned, a non-linear optical loop mirror as wavelength converter for achieving tunable delays [13], or slow-light delay-line basing on stimulated-Brillouin scattering [14]. But in these schemes, an extension to multiple wavelength-division-multiplexed (WDM) channels is challenging and polarization stabilization is usually required.

In this work, we propose and demonstrate the use of a commercially available programmable wavelength-selective switch (WSS) (Finisar WaveShaper 4000S) [15] to serve as an “all-in-one” platform that performs simultaneous demodulation and demultiplexing of NRZ-DPSK signals operating at different rates. We first study the bit error rate (BER) performance of the system using the WSS in terms of the receiver sensitivity and also the OSNR. We compare the results with the DLI receiver both at 10 Gb/s and 40 Gb/s. We then examine the system performance in a WDM scenario to demonstrate simultaneous demodulation and demultiplexing, and evaluate crosstalk. The WDM channels are separated by the ITU-grid standard of 100 GHz and operate simultaneously at 10 Gb/s and 40 Gb/s. We successfully demodulate four WDM channels after transmission in a fiber link and show error-free operation. The high flexibility of this approach makes it an ideal candidate for DPSK demodulation in future fiber networks requiring coexistence of multiple rates on the same transmission link. Basing on a comparative study with the DLI demodulation scheme, we show a similar-if not better- BER performance at higher rates.

2. Principle of operation

The 1×4 WSS consists of a 2D spatial light modulator (SLM) placed at the focal plane of a 4f Fourier transform processor; its operation has been described in greater detail in [15]. Briefly, using the x-axis pixels of the 2D SLM, the spectrum of a broadband signal can be controlled in amplitude and phase. The spatial phase in the y-axis direction provides an extra degree of freedom for routing spectral portions of the filtered signal to any of the four output fiber ports in the device. The synthesized spatial mask can be reconfigured by the user at an effective speed of several hundreds of ms, limited by the communication between the computer and the WSS. The key point for using this device as a DPSK demodulator is the possibility to implement an optical filter with a spectral response emulating the one obtained at the constructive port of a DLI for a determined transmission rate. After filtering, the selected channel is then demultiplexed and routed through the port defined by the user. A photodetector can then be used in a single-ended configuration.

The narrowest band-pass filter that can be synthesized with this model has a full-width at half maximum (FWHM) of 9.3 GHz, as illustrated in Fig. 1(a) . Since this is broader than the ideal pass-band filter of 6.67 GHz width for 10 Gb/s operation [3], we expect a slight penalty when compared to the DLI. At 40 Gb/s however, we synthesize the WSS to match three lobes of the spectral response of the DLI. A multi-lobe filter offers a better performance than single-lobe operation [5]. Three lobes are the maximum that can be implemented if the channels are to be spaced by the 100 GHz ITU grid standard.

 

Fig. 1 Dotted: measured spectral WSS filter response at (a) 10Gb/s and (b) 40 Gb/s. Solid line: the corresponding ideal DLI spectral response.

Download Full Size | PDF

3. Single channel comparative study

First, we assess the single-channel capabilities of the WSS for demodulating 10 Gb/s and 40 Gb/s NRZ-DPSK signals individually. The setup under consideration is shown in Fig. 2 . The DPSK transmitter consists of a continuous-wave (CW) laser modulated by a Mach-Zehnder modulator (MZM) biased at the minimum transmission point to provide NRZ-DPSK modulation [1]. We use a pseudo-random bit sequence (PRBS) with a length of 2³¹-1 bits as the data pattern. The comparative study is done in both back-to-back configuration and after propagation through 85 km of LEAF fiber. The average power launched to the transmission link is about 4 dBm at either of the transmission rates. For the 40 Gb/s experiments only, we use a length of dispersion compensating fiber (DCF) providing −300 ps/nm of dispersion. For back-to-back measurements, a variable optical attenuator (VOA) replaces the fiber link to ensure the same input power to the EDFA and therefore identical amount of amplified spontaneous emission (ASE) noise. The DPSK demodulator is either a DLI or the WSS programmed with the filter spectral response shown in Fig. 1(a) and Fig. 1(b) for 10 Gb/s and 40 Gb/s demodulation, respectively. We use a 40 GHz bandwidth photodiode for direct detection, and the detected sequence is sent to a bit-error-rate tester. We add ASE noise to the DPSK modulated signal in order to measure the BER performance with respect to the optical signal-to-noise ratio (OSNR) of the launched signal. The OSNR is measured with an optical spectrum analyzer (OSA) with 0.1 nm resolution for 10 Gb/s and 0.5 nm for 40 Gb/s. All the band-pass filters (BPFs) displayed in Fig. 2 are used to suppress the out-of-band ASE noise from the amplifiers.

 

Fig. 2 Experimental setup for the comparative study of DPSK demodulation for a single channel and single rate. BPF: band pass filter, BERT: bit error rate tester, DCF: dispersion compensating fiber, DMOD: demodulator, EDFA: erbium doped fiber amplifier, MZM: Mach Zehnder modulator, OSA: optical spectrum analyzer, PD: photodiode, and VOA: variable optical attenuator.

Download Full Size | PDF

At 10 Gb/s, the measured BER versus received average power is shown in Fig. 3(a) . The high power levels are due to the fact the photodiode does not have any amplification stages associated with it. We observe that at a fixed OSNR (32 dB), we get a 3-dB penalty in terms of receiver sensitivity when using the WSS. This penalty is due to the mismatch between the WSS and the DLI spectral response [see Fig. 1 (a)] as a result of its limited resolution. Figure 3 (b) illustrates the BER performance in the presence of added ASE noise. Here, we see a more significant OSNR penalty since the filter response implemented by the WSS is broader than that corresponding to the DLI, thereby allowing more out-of-band ASE noise to pass to the photodiode. Similar performance degradation in terms of Q-factor due to non-ideal pass-band filtering has also been reported previously [3]. No BER dispersion-induced penalty is observed neither for the DLI nor the WSS. Using a WSS with better resolution, we would expect to achieve a BER performance identical to the DLI. Nevertheless, we point out that it is still possible to achieve error-free operation even without DCF. Finally, in Fig. 3(c), we show the eye diagrams for the demodulated DPSK signals using the WSS and the DLI both in single-ended configuration. From the back-to-back configuration, we observe an open eye with different shapes depending on the demodulator [3]. A slight closure due to uncompensated dispersion is observed after propagation.

 

Fig. 3 Single channel 10 Gb/s DPSK demodulation results. (a) BER versus average received power, (b) BER versus OSNR, (c) the eye diagrams. DLI: delay line interferometer, WSS: wavelength selective switch.

Download Full Size | PDF

Figure 4(a) plots the BER versus the average received power at 40 Gb/s at a fixed OSNR of 32 dB. We observe a very similar performance between the WSS and the DLI. Comparing the back-to-back BER curves, there is a slight 1-dB improvement when using the WSS that we attribute to temperature and polarization instability issues in the DLI. Figure 4(b) shows the BER vs. OSNR. We again observe that the WSS offers a slightly better resistance to transmission impairments thanks to its stability to environmental and polarization variations. The eye diagrams of the demodulated signals using the DLI and the WSS are shown in Fig. 4(c), having a similar shape because of the accurate filter shape emulated by the WSS at 40 Gb/s. Since in this case dispersion is compensated for, we do not observe a distortion in the eye diagrams.

 

Fig. 4 Single channel 40 Gb/s DPSK demodulation results. (a) BER versus average received power, (b) BER versus OSNR, (c) the eye diagrams. DLI: delay line interferometer, WSS: wavelength selective switch.

Download Full Size | PDF

The experimental results show a better performance at 40 Gb/s compared to 10 Gb/s. This is due to the resolution of the device which does not allow for a bandwidth less than 9.3 GHz. Previous research papers have studied the impact of the filter bandwidth on the Q factor and shown a significant penalty when dealing with broad filters for NRZ-DPSK [3]. We have performed a numerical simulation to verify whether the corresponding OSNR penalty at 10 Gb/s is reasonable for this non-ideal filter. The simulation is performed with the commercial software OptiWave^TM and the parameters are selected to match the settings of our experiment (except for the photodiode and the BER tester, which requires precise noise and bandwidth characterization) for single channel operation in a back-to-back configuration. We assume a single CW laser, modulated by an MZM biased at the minimum transmission point, and mixed with ASE noise filtered in the same band as the CW signal to control the OSNR, which is measured by a simulated OSA with 0.1 nm resolution, as in our setup. The demodulator uses either a DLI with the same FSR as used in our experiments or a narrow-band spectral Gaussian filter, both in a single-ended detection configuration.

Figure 5(a) shows the simulated penalty in terms of received power for DLI and the Gaussian filter with different FWHM widths. As expected, the DLI offers the best performance; there is an optimum bandwidth for the filter (between 6 and 7 GHz). This is consistent with the results reported in [3] about the optimum filter bandwidth. The simulation shows a penalty slightly worse than 1 dB, which is in reasonable agreement to the measured 3 dB. For these measurements a fixed OSNR of 32.7 dB is used. Figure 5(b) shows the simulated BER curves vs. OSNR for a fixed received power (−14 dBm). The optimum filter bandwidth is around 7 GHz for which the penalty is minimized. We observe an 8-dB penalty for the filter bandwidth similar to the one used in the experiment, which is in close agreement to the 10 dB measured penalty. The simulations confirm that the penalty for the band-pass filter as compared with the DLI is due to the finite resolution of the WSS. The slight discrepancies between the measured and simulated BER penalties might be due to different photodiode thermal noise and/or the BER tester operation and noise. Precise characterization and modeling of these components is required for accurate comparison between experimental measurements and the simulation results, which is beyond the scope of this work.

 

Fig. 5 (a) Simulation results for the single channel at 10 Gb/s. (a) BER versus average received power, (b) BER versus OSNR.

Download Full Size | PDF

4. Simultaneous multi-rate WDM demodulation and demultiplexing

In the previous Section, we showed that is possible to demodulate an NRZ-DPSK signal operating either at 10 Gb/s or 40 Gb/s using the programmable WSS and obtain error-free operation. We now show that the same device can demodulate and demultiplex simultaneously several WDM channels operating at different rates. We implement the spectral filter shape corresponding to the bit rate of the particular wavelength of interest and route each filtered channel through a specific output port of the device. The setup is illustrated in Fig. 6 (a) . Four CW lasers, spaced by 100 GHz, are used as light sources. Two of them are modulated by an MZM driven by a 10 Gb/s 2³¹-1 PRBS signal, and the other two by an independent 40 Gb/s 2³¹-1 PRBS sequence using a second MZM. Both MZMs are biased at the minimum transmission point to produce the desired NRZ-DPSK format. The four channels are coupled and transmitted through 85 km of the LEAF fiber and DCF. The spectrum of the signals before transmission is shown in Fig. 6(b). The synthesized filter shapes corresponding to ports 4 and 3 are shown in Fig. 6(c) and Fig. 6(d), respectively. We observe leakage from adjacent channels when all the WSS channels are in operation (dashed-blue line) compared to the case obtained when only that particular channel is on and the rest are blocked with the aid of the SLM (green solid line). Similar results are obtained for channels 1 and 2. Thus, one might expect some receiver penalty due to cross-talk.

 

Fig. 6 (a) Experimental setup for demodulation of multi-rate WDM DPSK signals, (b) spectrum of the signals before transmission, (c) synthesized filter shape for 10 Gb/s demodulation at port 4, and (d) for 40 Gb/s demodulation at port 3, solid: the rest of the ports are blocked, dash-dot: all the ports are on.

Download Full Size | PDF

The measured BER versus average received power obtained for simultaneous demodulation of the four channels are presented in Fig. 7(a) and Fig. 7(b) for 10 Gb/s and 40 Gb/s, respectively. At 10 Gb/s, as shown in Fig. 7(a), a 2-dB penalty is observed from one channel to another, which we attribute in part to the wavelength dependence of the MZM transmission. In a real fiber transmission system, independent modulators at each channel are used instead, and therefore their settings are individually optimized to alleviate this penalty. More importantly, the cross-talk induced penalty is less than 1 dB, indicating that the spectral filtering leakage effect seen in Fig. 6 has a minor impact on the receiver sensitivity. At 40 Gb/s, as shown in Fig. 7(b), the performance is almost identical for all individual channels, regardless of the operation in the rest of the ports. These results indicate that the cross-talk induced penalty at the WSS is negligible in terms of receiver sensitivity.

 

Fig. 7 BER versus average received power for multi-rate WDM DPSK demodulation at (a) 10 Gbs and (b) 40 Gb/s.

Download Full Size | PDF

5. Summary and conclusions

We have used a programmable WSS to implement band-pass filtering to demodulate NRZ-DPSK signals in a single-ended configuration. This scheme allows us to perform simultaneous demodulation and demultiplexing of four WDM channels after transmission in a dispersion-compensated fiber link. Owing to the finite resolution of the WSS, a better performance at 40 Gb/s is obtained. Error-free demodulation for all channels is successfully achieved at two rates (10 Gb/s and 40 Gb/s). Given the spectral resolution and robustness of the system, this scheme offers interesting perspectives for future fiber optical networks where the coexistence of multiple channels and multiple ultrahigh-repetition rates (including 40 Gb/s and beyond) will be required to handle simultaneously the increasing demands of bandwidth and traffic diversity.