1. Introduction

All-optical logic operation is expected to play important roles in future high-speed optical networks where reducing latency and power consumption is of great interest. For example, all-optical header processing based on optical logic can mitigate the processing speed bottlenecks and reduce high power consumption in optical data networking owing to the limitations in high-speed electronic signal processing [1, 2]. Among the family of Boolean logic functions, Exclusive-OR (XOR) plays central roles in all-optical bit-pattern matching [3,4], all-optical label swapping [5], parity checking [6], and pseudo-random bit sequence generation [7].

All-optical XOR operation of on-off-keyed (OOK) optical signals can be implemented utilizing cross-phase and gain modulation in a nonlinear optical medium in combination with an optical interferometer; for example, a Mach-Zehnder interferometer (MZI) incorporating semiconductor optical amplifiers (SOAs) [4, 5, 8, 9], SOA-based ultrafast nonlinear interferometers (UNIs) [10–12], and fiber-based nonlinear optical devices [13, 14]. It is well known that the speed of SOA-based optical signal processing is ultimately limited by the penalties arising from the nonlinear patterning effects due to the non-instantaneous recovery dynamics of the cross-phase and gain modulation. The highest bit-rates at which XOR using SOAs has been demonstrated for pseudo-random OOK signals are 40 Gb/s using the conventional SOA-MZI configuration [8, 9], and 85 Gb/s using “turbo-switch” in a double-UNI configuration [12]. In “turbo-switch” configuration, SOAs are cascaded where the SOAs in the latter stages act as nonlinear temporal filters to reduce the patterning effect. For example, three SOAs in series are used to achieve 85 Gb/s XOR in [12].

Here, we propose an alternative method of mitigating the patterning effects and achieving increased-speed all-optical XOR operation that is well suited for large scale integration of cascadable logic operations. The method utilizes a photonic-integrated SOA-MZI pair each differentially driven by a data input and its complement. We demonstrate simultaneous XOR and XNOR operations at 86.4 Gb/s using SOAs having a carrier recovery time of ~20 ps.

2. Principle

The proposed circuit accomplishing simultaneous all-optical XOR and XNOR operations is shown in Fig. 1 . It consists of a pair of SOA-MZIs. Each SOA-MZI is differentially driven by data and complement. Note that a single SOA-MZI can be used for all-optical XOR as previously demonstrated in [8, 9]. The arrangement in Fig. 1, in contrast to the conventional single SOA-MZI XOR set up, ensures that each SOA receives a nearly constant-power stream of input pulse train, except for the slight temporal shifts due the differential input delay t. This is illustrated in Fig. 1 using an example of an on-off-keyed (OOK) signal A=[101]. Hence the patterning effects can be mitigated that arise from the fluctuation in the signal optical powers when SOAs are directly modulated by OOK signals without the differential set up. For instance, the most deleterious situation is avoided, where the SOAs receive a long sequence of 1's followed by a long sequence of 0's, or vice versa.

 

Fig. 1 Schematic of the all-optical XOR circuit. CW: continuous wave laser. τ: differential time delay.

Download Full Size | PDF

The working principle of the device shown in Fig. 1 is as follows: the top SOA-MZI converts the input OOK data streams A and its complement $\overline{\text{A}}$ into a binary phase shift keyed (BPSK) signal, exp(jπA). The bottom SOA-MZI similarly outputs a BSPK signal exp(jπB). The conversion of OOK into BPSK was first used for all-optical wavelength conversion of DPSK signals [15] and its process is detailed in [9, 15]. The next step in XOR operation is the linear optical interference between these two BPSK signals. The intensity envelopes of the output signals emerging from the constructive and destructive ports of the 2x2 coupler are calculated using:

The proposed method enhances XOR operation speed by suppressing the patterning effects but does require the availability of complementary data input. However, the added complexity is mitigated by the following factors that are favorable for scalable integration of multiple optical logic gates: firstly, the circuit provides both XOR and XNOR output data, which can be used as inputs for the next stages of additional XOR operations. Secondly, in larger scale integration of multiple logic gates, amplification of signals using SOAs is required between each steps of operation, for which PSK in comparison to OOK has the advantage of reduced patterning effects during the SOA amplification. Thus, optical amplification can be performed at locations where the signal of interest is in a PSK format.

3. Experiment

We implement the circuit outlined in Fig. 1 using a hybrid photonic integrated device incorporating InGaAsP quantum well SOAs and a silica planar light wave circuits (PLCs). We show in Fig. 2 a picture of the packaged device. The device consists of three parts: the passive waveguides making up the input and output interferometer parts are implemented using 4% index-contrast silica-on-silicon PLCs. The high-index contrast helps in reducing the device footprint. The optical couplers in the silica PLC are implemented using 1x2 and 2x2 multi-mode interferometers (MMIs). The array of four semiconductor optical amplifiers is a commercial device from CIP Technologies (Quad-SOA-NL-1550) with 20-25 ps saturated SOA gain recovery time. The overall size of the device is 2.2 cm x 0.5 cm. The silica PLCs and SOAs are actively aligned by maximizing the coupled amplified spontaneous emission (ASE) from the SOAs into the silica PLCs. Using spot size converters on the SOAs and also on the PLCs, coupling loss between the SOAs and the silica PLC is less than 1 dB per facet.

 

Fig. 2 A picture of the packaged hybrid XOR device.

Download Full Size | PDF

In Fig. 3 , we show a schematic of the set up to obtain the OOK data signals using optical time division multiplexing (OTDM): we start with two commercial 10.8-GHz semiconductor mode-locked lasers (U2t TMLL 1550) as the pulsed sources producing ~ 2 ps pulses. Pulse trains from each laser is 4:1 multiplexed via fiber delay lines into two 43.2-GHz pulse trains, of which one is modulated by a 43.2-Gb/s pseudo-random bit sequence (PRBS) D of length 2³¹-1, using an electroabsorption modulator (OKI OM5642W-30B), and the other is modulated by the complementary data $\overline{\text{D}}$ using a similar electroabsorption modulator. These data streams are further multiplexed to obtain 86.4-Gb/s data streams A($\overline{\text{A}}$) and time-delayed copies B($\overline{\text{B}}$). In practice, complementary data input are not always available and need to be generated from the data input. One can exploit an all-optical inversion circuit, which can be implemented using a SOA-MZI, to obtain the complementary data stream. In our demonstration, such a scenario is simulated by each MLL having a different wavelength. (1560 nm and 1557 nm). The wavelength of CW is 1550 nm. The power of CW launched to the XOR circuit is 14 dBm and the power of each data and complementary data input into the device is ~11 dBm. We estimate that the optical pulse energy coupled to the SOAs is ~12 fJ.

 

Fig. 3 A layout of the OTDM set up to generate input signals to the XOR circuit. ODL: optical delay line. MLL: mode-locked laser.

Download Full Size | PDF

We adjust the hybrid XOR device first by optimizing the BPSK signals from each of the MZI pair by monitoring the optical spectra and the eye diagrams measured using a 50-GHz photodiode connected to an 80-GHz sampling module. After BPSK optimization, XOR and XNOR operations are validated using sampling scope traces. We show an example of XOR and XNOR operations in Fig. 4 , where it is straightforward to verify the correct logic operations keeping in mind the fact that XOR is equivalent to binary addition modulo two. We show the signal quality of XOR operation using the eye diagram shown in Fig. 5 . The observation of open eye diagrams for 86.4 Gb/s XOR operation is a substantial improvement over our previous 80 Gb/s XOR operation using a conventional SOA-MZI configuration [9], with which we failed to achieve an open eye diagram. The recovery speed of the SOAs used in [9] is nearly identical to that of the SOAs used for this demonstration, and the result shows the benefit of reduced patterning effects using the proposed method.

 

Fig. 4 Temporal traces of input data A, B and the XOR and XNOR results.

Download Full Size | PDF

 

Fig. 5 Eye diagrams of simultaneous XOR (left) and XNOR (right) operations.

Download Full Size | PDF

As can be seen in Figs. 4 and 5, the signal quality, more specifically the extinction ratio, of XNOR is inferior to that of XOR. The causes of the signal degradation are mainly two fold: firstly, the uneven splitting ratios of the 1x2 MMIs render the MZIs asymmetric, the result of which is incomplete cancellation of CW component in the BPSK signals. Secondly, the incomplete extinction of the BPSK signals is further magnified in the constructive interference process in generating XNOR signal output, while it is canceled in the destructive interference for generating XOR output.

Aside from correcting the aforementioned device imperfections, we plan to further optimize the device. It has been reported that the all-optical signal processing using differential SOA-MZI can greatly benefit from optimization of the differential delays and the power splitting ratio of the differential input stage [16]. In the current device, the differential delay is fixed to 3 ps and the nominal splitting ratio is 50%, both of which can be optimized. Obviously, SOAs with faster recovery speed such as reported in [17] will allow faster XOR/XNOR operation at higher speeds.

4. Summary

We proposed a novel method of all-optical XOR operation using SOAs with the capability of suppressing the patterning effects owing to the slow carrier recovery speed of the SOAs. We fabricated a hybrid photonic integrated device consisting of passive silica PLC and an array of four SOAs, implementing a pair of SOA-MZIs. We demonstrated a high-quality all-optical XOR operation at 86.4 Gb/s using the hybrid photonic integrated device.