1. Introduction

All-optical signal processing techniques are highly desirable in future large-capacity optical networks for the high processing speed and no need for optical-electrical-optical (O-E-O) conversions [1]. Optical wavelength multicasting and optical logic gates are essential functions of optical signal processing. In the optical networks today, migration operation in data centers, transferring the user data from one data center to others where different wavelengths are employed, is a challenging issue. With optical wavelength multicasting technologies, a huge amount of user data of emerging applications like streaming media, high-definition TV can be efficiently delivered from the sever side to the user sides. The wavelength multicasting scheme makes the migration operation more flexible and efficient [2], and it is becoming an essential technology for data migration between data centers in the future optical networks. On the other hand, all-optical logic gates are useful network elements in addressing, switching, header recognition, data encoding, regeneration, parity checking [3, 4] and network coding [5, 6]. Ultrafast logic gate operations in the optical domain can potentially enable digital-signal-processing functions at a high-speed transmission line rate such that network latency can be decreased with improved performance. In the last decades, with the ever-growing Internet traffic, high spectrum-efficient phase-modulated signals attracted increased interest for use in long-haul, high-capacity WDM systems. It is highly desirable to investigate wavelength multicasting and optical logic gate operation schemes for advanced phase modulated modulation formats.

Previously, various wavelength multicasting schemes have been studied and demonstrated for differential phase-shift keying (DPSK) signal in various nonlinear devices including highly nonlinear fiber (HNLF) [7–9], Silicon waveguide [10] and semiconductor optical amplifier (SOA) [11]. Optical logic gates, including half-adder, full-adder, and exclusive-OR (XOR), could be constructed using all-optical processing approaches [12]. In the past few years, logic XOR gate for DPSK signal has also been demonstrated using highly nonlinear fiber (HNLF) [13, 14], SOA [15, 16] or periodically poled lithium niobate (PPLN) waveguide [17]. However, up to now, all the reported works of wavelength multicasting aimed at delivering the information carried by one input wavelength to several different output wavelengths, i.e., single-channel wavelength multicasting, SWM. It will be interesting to simultaneously multicast multiple input signals, namely delivering the information carried by two or more input wavelengths carrying different data to different destination wavelengths, i.e., multichannel wavelength multicasting, MWM. By employing one MWM module, individual SWM modules could be replaced, resulting in compact system design and lower power consumption. With MWM scheme the capacity and flexibility of the optical network could be further improved. In addition, most of the reported works of multicasting have the expenditure of additional pumps, which increases the implementation complexity and cost. On the other hand, the XOR logic operation of DPSK signal was achieved in [15] and [16] with only one-channel XOR output. Multi-channel output XOR for two-input DPSK signals in [13, 14] was obtained, but it required additional pump light. XOR multicasting operation for more than two input DPSK signals without pump light’s participation would be attractive. Moreover, so far, wavelength multicasting and optical logic gates in the previous reported works were performed separately. A laudable goal is to realize these important network functionalities simultaneously using single nonlinear element. For example, in the future optical network nodes, it will be attractive and efficient to simultaneously implement multiple wavelength multicasting to avoid channel contention or delivering data streams to numerous customers that employing different wavelengths, and logical gate operation to perform switch and control.

Recently, quantum-dot SOA (QD-SOA) has attracted considerable interest for its unique properties like higher gain, faster response, lower noise figure (NF) and broader gain bandwidth compared with the traditional bulk/quantum-well SOAs. Multicasting and logic operation have been separately demonstrated using cross-gain modulation (XGM) or four-wave mixing (FWM) effects [18–20] in QD-SOA. In this paper, to the best of our knowledge, for the first time, we propose and experimentally demonstrate simultaneous MWM and XOR logic gate multicasting (XOR-LGM) for three input NRZ-DPSK channels based on FWM in QD-SOA. No additional pump light is needed in the scheme. Through the interaction of the input three lights, each channel is successfully multicasted to three wavelengths (1-to-3 for each), totally 3-to-9 MWM, and at the same time a three-output XOR-LGM is obtained at three different wavelengths. Error-free operations were achieved for all of the obtained signals with a power penalty less than 1.2dB at a bit-error rate (BER) of 10⁻⁹.

2. Operation principle

The proposed simultaneous MWM and XOR-LGM scheme for three input NRZ-DPSK channels is schematically illustrated in Fig. 1. The scheme is based on three lights’ FWM, with no additional pump light’s participation. As shown in the Fig. 1, there are three input DPSK signals as FWM participators in SOA: DPSK1 at wavelength${\lambda }_1$, DPSK2 at wavelength ${\lambda }_2$ and DPSK3 at wavelength${\lambda }_3$. The frequency spacing of DPSK1-to-DPSK2 and DPSK2-to-DPSK3 is $\Delta f_1$ and$\Delta f_2$, respectively. Both degenerate FWM (D-FWM) and non-degenerate FWM (ND-FWM) processes happen inside the QD-SOA with new frequencies generated [21], as schematically illustrated in Fig. 1. Each of the generated components through D-FWM and ND-FWM possesses a frequency of $f_{abc}=f_a+f_b-f_c$(b≠c, a b and c $\in$ 1, 2, and 3), and a phase of ${\phi }_{abc}={\phi }_a+{\phi }_b-{\phi }_c$(b≠c, a b and c $\in$ 1, 2, and 3) [16, 22]. The components with a = b correspond to D-FWM components while the ones with a≠b are ND-FWM components. Finally nine idlers are generated, which are indicated by ${\lambda }_{1-a}$~${\lambda }_{1-b}$, ${\lambda }_{2-a}$~${\lambda }_{2-b}$, ${\lambda }_{3-a}$~${\lambda }_{3-b}$and ${\lambda }_{XOR1}$~${\lambda }_{XOR3}$ in Fig. 1. The relationship of the electrical field ($E$) and the optical phase ($\phi$) for the generated idlers at ${\lambda }_{1-a}$~${\lambda }_{1-b}$, ${\lambda }_{2-a}$~${\lambda }_{2-b}$, ${\lambda }_{3-a}$~${\lambda }_{3-b}$ (D-FWMs) and ${\lambda }_{XOR1}$~${\lambda }_{XOR3}$ (ND-FWMs) can be expressed as [23]:

 

Fig. 1 Operation principle of the simultaneous MWM and XOR-LGM scheme for three input NRZ-DPSK signals based on four-wave mixing without additional pump in QD-SOA.

Download Full Size | PDF

Table 1. Phase Patterns of the Simultaneous MWM and XOR-LGM Scheme for Three Input NRZ-DPSK Channels

View Table | View all tables in this article

3. Experiment setup and results

The experimental setup of our proposed simultaneous MWM and XOR-LGM scheme for three input NRZ-DPSK channels is presented in Fig. 2. Three polarization maintaining tunable lasers (PM-TL1~3), a polarization maintaining coupler (PM-OC) and one inphase/quadrature (IQ) modulator are used to generate the input three 10Gbps NRZ-DPSK signals. One arm of the IQ modulator is driven by an amplified electrical data stream from a pulse pattern generator (PPG, Anritsu MP1761B) having a pseudorandom bit sequence (PRBS) with the bit pattern length of 2¹⁵−1. The other arm of the IQ modulator is left un-modulated. The wavelengths of the three signal lights are set as: ${\lambda }_1$ at 1548.689nm, ${\lambda }_2$ at 1549.494nm and ${\lambda }_3$ at 1551.902nm. After generation, an erbium-doped ðber ampliðer (EDFA) is used to amplify the signals. After the EDFA, the three input signals are separated using an arrayed waveguide grating (AWG), and then delayed by integral bit periods with each other using two optical delay lines (ODL1 and ODL2). The three lights are set to be co-polarized with each other by properly adjusting the polarization controllers (PCs) to reduce the system polarization sensitivity [11, 24-25] and obtain the highest FWM conversion efficiency in QD-SOA. All the three input signals are recombined together through an OC and then are injected into the QD-SOA. The power of the three input DPSK signals is kept even, and the total power injected to the QD-SOA is 5dBm. The two circulators before and after the QD-SOA are used as isolators to decrease the influence of the reflection. In the experiment, the bias current and temperature are set as 500 mA and 25°C, respectively. At the output of the QD-SOA, the new generated FWM idlers are filtered by an optical band-pass filter (OBPF) with variable central wavelength. The filtered idlers are then amplified by two cascaded EDFAs. A delay line interferometer (DLI) and a balanced photo-diode (BPD) are used to demodulate the signal and convert the signal into electrical signal. The bit-error-rate (BER) of the signal is measured using an error detector (ED, Anritsu MP1762A).

 

Fig. 2 Experiment setup. PM-TL: polarization maintaining tunable laser; PM-OC: polarization maintaining optical coupler; EDFA: erbium-doped ðber ampliðer; AWG: arrayed waveguide grating. ODL: optical delay line. PC: polarization controller. OBPF: optical band-pass filter. ATT: attenuator. DLI: delay-line interferometer; BPD: balanced photo detector;

Download Full Size | PDF

The QD-SOA that we fabricated contains highly stacked Stranski-Krastanow QDs [26] and has a 2-mm-long device length. The gain of the device is dominated by transverse electric (TE) mode. Maximum 25 dB gain is measured around C-band at a bias current of 400 mA [27]. The complete (100%) carrier recovery time of the QD-SOA is about 30ps. FWM efficiencies of higher than −40 dB were achieved within a 23-nm range. Details of the device that we used are given in [27].

Figure 3 shows the optical spectra measured by an optical spectral analyzer (OSA, ANDO AQ6317B) at the input and output of the QD-SOA. In the measured optical spectra, the three input DPSK signals are indicated by${\lambda }_1$, ${\lambda }_2$ and${\lambda }_3$, respectively. The new generated components are marked by${\lambda }_{1-a}$,${\lambda }_{1-b}$,${\lambda }_{2-a}$,${\lambda }_{2-b}$,${\lambda }_{3-a}$,${\lambda }_{3-b}$and ${\lambda }_{XOR1}$~${\lambda }_{XOR3}$. As analyzed above, ${\lambda }_{1-a}$ ~${\lambda }_{1-b}$, ${\lambda }_{2-a}$ ~${\lambda }_{2-b}$and ${\lambda }_{3-a}$ ~${\lambda }_{3-b}$ preserve the information of the original input DPSK1, DPSK2 and DPSK3, respectively. And the components at ${\lambda }_{XOR1}$~${\lambda }_{XOR3}$are the logic XOR gate of the input three signals. The spectrum indicates that all the components of MWM and XOR-LGM are successfully obtained after FWM. Among the newly generated components, ${\lambda }_{1-a}$ ~${\lambda }_{1-b}$, ${\lambda }_{2-a}$ ~${\lambda }_{2-b}$and ${\lambda }_{3-a}$ ~${\lambda }_{3-b}$ have conjugated phase with respect to the input DPSK1, DPSK2, and DPSK3, respectively. Optical phase conjugation is one of practical approaches to compensate for both chromatic dispersion and nonlinearity in long-haul transmission. With our current experimental setup, by tuning the frequency spacing between the input three signals, the output MWM and XOR-LGM channels could be consistent with the ITU channels. In the experiment, it is better to select the wavelengths of the input signals around the gain peak at 1533nm to get a high gain for all the signals. However, due to the wide gain bandwidth of the QD-SOA, the choice of the wavelengths could be flexible. To avoid frequency overlapping of the new generated components, the minimum frequency space of ${\lambda }_2$-to-${\lambda }_3$($\Delta f_2$) should be larger than $2\Delta f_1$ ($\Delta f_2>2\Delta f_1$).

 

Fig. 3 Optical spectrum at the (a) input and (b) output of the QD-SOA.

Download Full Size | PDF

To characterize the performance of the proposed simultaneous MWM and XOR-LGM scheme for three input NRZ-DPSK channels, BER curves versus the received power for each output channel is measured and shown in Fig. 4. The back-to-back (BTB) performance of the three input DPSK signals are also presented as references. The results in Fig. 4 indicate that error-free operations are obtained for the converted channels. Compared to the BTB case, the maximum power penalty of the output wavelength multicasting channels is 1.2dB. And for the three-output logic XOR gate, the maximum power penalty is 0.6dB.

 

Fig. 4 Measured BER performance versus received power for the output MWM and XOR-LGM channels.

Download Full Size | PDF

Here, the conversion efficiency (CE) is defined as the ratio of the converted signal power to that of the input signal. The detailed performances including the wavelengths, the CE and the power penalty at BER of 10⁻⁹ of the output MWM and XOR-LGM channels are summarized in Table 2. The corresponding eye diagrams of the input three DPSK signals and the output MWM and XOR-LGM channels after demodulation are shown in Fig. 5. The clear and open eyes prove the good quality of the output signals.

Table 2. Performance of the Simultaneous MWM and XOR-LGM Scheme

View Table | View all tables in this article

 

Fig. 5 Measured eye diagrams at the output of the QD-SOA for (a)${\lambda }_1$, 1548.689nm. (b) ${\lambda }_{1-a}$,1550.298nm. (c) ${\lambda }_{1-b}$,1555.119nm. (d) ${\lambda }_{XOR1}$,1546.291nm. (e) ${\lambda }_2$,1549.494nm. (f) ${\lambda }_{2-a}$,1547.890nm. (g) ${\lambda }_{2-b}$,1554.316nm. (h) ${\lambda }_{XOR2}$,1551.099nm. (i) ${\lambda }_3$,1551.902nm. (j) ${\lambda }_{3-a}$,1545.495nm. (k) ${\lambda }_{3-b}$,1547.092nm. (l) ${\lambda }_{XOR3}$,1552.705nm. All with 20ps/div.

Download Full Size | PDF

To further verify the logic XOR gate multicasting operation, the temporal waveforms (bit patterns) are captured for different optical waves. Figure 6 and Fig. 7 show the temporal waveforms of demodulation outputs from the DLI for the input three NRZ-DPSK signals and the output three channels at ${\lambda }_{XOR1}$~${\lambda }_{XOR3}$. It can be clearly seen that all the three demodulated idler waves satisfy the XOR logic operation for the three input DPSK signals. The slight degradation of the XOR outputs may be attributed to the phase and intensity noise in FWM process.

 

Fig. 6 50-bit demodulated waveforms of the three input signal: (a) DPSK1, (b) DPSK2, (c) DPSK3.

Download Full Size | PDF

 

Fig. 7 50-bit demodulated waveforms of three XOR-LGM channels at (a)${\lambda }_{XOR1}$, (b)${\lambda }_{XOR2}$, (c)${\lambda }_{XOR3}$.

Download Full Size | PDF

Note that, with the current QD-SOA sample, to avoid breaking the device, the bias current and the launched total power are limited to 500mA and 5dBm, respectively. In the future, when fabricating QD-SOA device, we will optimize the linewidth of the waveguide and thickness of the waveguide layer to enable the QD-SOA to handle higher optical launch power and bias current, thus improving the conversion efficiency and performance of the wavelength multicast based on the FWM in QD-SOA. An extension of this scheme in future can be used for simultaneous MWM and XOR-LGM for more than three DPSK channels without the use of additional pumps. However, when the input channel is more than three, the frequency detuning of the input lights needs to be adjusted properly to avoid frequency overlap for the new generated idlers, frequency overlap will cause severe crosstalk [7]. On the other hand, higher order phase modulated signals like quadrature phase-shift keying (QPSK) can also be considered in this simultaneous MWM and XOR-LGM scheme, but additional pumps may be needed. Owing to the ultra-fast characteristic of QD-SOA, it also can support signal processing for higher bit rate signals.

4. Conclusion

Simultaneous multichannel wavelength multicasting and XOR logic gate multicasting for three input NRZ-DPSK channels based on FWM without any additional pump lights’ participation in QD-SOA has been discussed and experimentally demonstrated. Through the interaction of the input three signals, each channel is successfully multicasted to three wavelengths (1-to-3 multicasting for each), totally 3-to-9 MWM, and at the same time a three-output XOR-LGM is obtained at three different wavelengths. The generated nine wavelength multicasting channels in 3-to-9 MWM are with a maximum power penalty 1.2dB at a BER of 10⁻⁹, and for the three XOR logic gate outputs, the penalty is less than 0.6dB at a BER of 10⁻⁹. The three-input three-output XOR has potential to be used in the optical network coding, which helps to protect the optical networks against network failures and thus avoid the loss of huge amount of user data [6]. The wavelengths of the generated MWM and XOR-LGM channels could be consistent with the ITU standard by tuning the frequency spacing between the input three signals. The QD-SOA is useful component for achieving multiple data and logic multicasting in the future transparent optical networks.