1. Introduction

In future high-speed optical networks it is crucial to develop all-optical signal processing to avoid the limits of electronic bottlenecks and increase the bit rate for each wavelength. All-optical logic gates are widely applied in all-optical signal processing, such as all-optical demultiplexing, switching, buffering, regenerating, and computing. Based on the nonlinearities of semiconductor optical amplifiers (SOAs), various logic gates have been reported in the literature. Schemes based on cross-phase modulation (XPM) by use of Sagnac gates[1], ultrafast nonlinear interferometers (UNIs)[2], Mach-Zehnder interferometers[3,4], or Michelson interferometers[5] are fit for higher-speed operation but depend on accurate phase control. Schemes based on cross-gain modulation (XGM)[6–8] usually employ more than one SOA. Schemes based on cross-polarization modulation [9,10] require equal polarization and matching wavelengths of the two input signal beams, which is quite impractical. A scheme based on four-wave mixing (FWM)[11] is limited by the FWM conversion efficiency. Hamie et al.[7] have achieved a logic NOR gate based on XGM of SOA at 2.5 GHz. However, this scheme requires two SOAs and an additional continuous-wave (cw) light besides the two input signal beams. Recently, a configuration of SOA-fiber ring laser (SFRL) was reported to be able to implement tunable, all-optical wavelength conversion at 2.5 Gb/s[12], all-optical clock recovery at 30 Gb/s[13], and ultrafast mode-lock pulses up to 40 GHz[14].

In this letter, we present a novel scheme for a tunable, all-optical NOR gate at 10 Gb/s based on XGM and an SFRL. It allows simple and efficient implementation because the logic operation only depends on the optical power rather than the optical phase that will need accurate phase control. Moreover, it requires only a single SOA and two input signal beams without additional light source, which makes this scheme quite attractive. The principle of operation is described in Section 2. Experimental results and corresponding discussions are presented in Section 3. Final conclusions are given in Section 4.

2. Principle of operation

 

Fig. 1. (a) Schematic diagram for NOR gate. (b) Principle of operation. (c) Corresponding truth able.

Download Full Size | PDF

The schematic diagram for the all-optical NOR gate is shown in Fig. 1(a). Signal A and signal B are coupled together and used as the pump beam, while a cw light is produced in the ring cavity and used as the probe beam. Knowledge of the output pulse power of the SFRL versus input pump pulse power is needed to clarify the principle of operation. However, it is not simple to measure the dynamic characteristics. Instead, the average output power versus average pump (periodic 1010 signal at 10 Gb/s) power at node a is measured to indicate the similar dynamic characteristics, as shown in Fig. 1(b). As long as one of the two input signals is bit “1,” including “01,” “10,” and “11,” the pump beam saturates the gain of the SOA deeply at low level, so the output signal is bit “0.” Once signal A and signal B are both bit “0” no pump beam exists, and the carrier density of the SOA recovers to a high level so the output signal turns to bit “1.” Therefore, Boolean NOR operation can be achieved. The corresponding truth table is shown in Fig. 1(c).

It should be noted that the average output power of the SFRL is not the same if the pump beam is changed to cw light [see Fig. 1(b)]. In this instance output power decreases almost linearly as pump power increases, and the laser will die when pump power becomes higher than 4 mw. However, this will not happen with a periodic 1010 pump signal because the carrier density of the SOA could recover some during the pump “0” period so as to always stay above the threshold level for lasing. Moreover, the pump beam and output signal do not propagate around the ring cavity, so that the response time for logic operation is limited only by the XGM effect of the SOA instead of depending on the time required to establish lasing in the ring cavity.

3. Experimental results and discussions

A. Setup and Results

 

Fig. 2. Experimental setup for logic NOR gate.

Download Full Size | PDF

The experimental setup is shown in Fig. 2. A NRZ signal at 1549.3 nm is generated by a bit-error-rate-test (BERT) system which is modulated by an external tunable-frequency synthesizer (TFS). BERT output optical power is boosted by an erbium-doped fiber amplifier (EDFA) and then attenuated by a tunable attenuator to adjust the optical power at node a. The output signal from node a is then divided into two channels through a 50:50 optical coupler C1. The optical beam in one channel is called signal A after passing a polarization controller (PC). A variable optical delay line ODL1 is inserted into the other channel, and the output beam from this channel is called signal B. Signal A and signal B are then coupled by the second 50:50 optical coupler C2 to produce combined signals (as pump beam) at node b. The laser as probe beam is produced in the fiber ring cavity. The SFRL consists of a SOA, an isolator, a fiber Fabry-Perot tunable filter (FFP-TF), the second variable optical delay line ODL2, and two couplers. Optical coupler C3 is used to introduce the pump beam into the ring cavity. The SOA is fabricated of 450 µm strained, multiquantum well (MQW), InGaAsP/InP material and has the characteristics of polarization independence. The gain peak wavelength is ~1550 nm, and the small signal fiber-to-fiber gain is 18 dB with a bias current of 100 mA. An isolator is employed here to allow the clockwise resonance only. The FFP-TF with a FWHM bandwidth of 1.6 nm is used for wavelength selection and also to prevent the pump beam from circulating in the cavity. With the FFP-TF tuned, a slight change in the total length of the ring cavity will result, and the ODL2 is used to compensate for this slight length change. Finally, 10% of the optical power is coupled out from the ring cavity by 90:10 optical coupler C4. Another 90:10 optical coupler C5 follows to distribute the output optical power to be measured by communication signal analyzer CSA and an optical spectrum analyzer OSA.

Output optical spectra involving two major lines are observed by the OSA. One at 1549.32 nm represents the pump signal which is retained in the ring cavity. The other line at 1557.84 nm represents the internal laser signal. The polarization independence of the SOA reduces polarization competition in the ring cavity; thus output optical spectra remain very stable.

The input and output bit streams at 10 Gb/s are observed by the CSA, as shown in Fig. 3. The waveform in Fig. 3(a) represents bit streams for signal A, periodic 1000110011101111, for which the optical power is 4.1mw for bit “1” (measured before C2, see Fig. 2). The waveform in Fig. 3(b) represents bit streams for signal B, periodic 1100011001110111, which are one bit period delayed compared with 1000110011101111, and the optical power is 4.3 mw for bit “1” (measured before C2). The waveform in Fig. 3(c) represents the combined result of signal A and signal B (measured at node b, see Fig. 2). It has three power values: P₁₁=4.4 mw when both A and B are bit “1,” P₁₀=2.1 mw when signal A (or B) is bit “1” while signal B (or A) is bit “0,” and P₀₀=75 µw when both A and B are bit “0.” The waveform in Fig. 3(d) represents output bit streams, periodic 0011000100000000 that are just the Boolean NOR between signal A and signal B. Output optical power is 241.6 µw for bit “1” and 15.4 µw for bit “0,” so the corresponding extinction ratio is 11.9 dB.

 

Fig. 3. Experimental results for NOR at 10 Gb/s.

Download Full Size | PDF

(a) Bit streams for signal A: periodic 1000110011101111.

(b) Bit streams for signal B: periodic 1100011001110111.

(c) Combined result of the two input signals.

(d) Output bit streams: periodic 0011000100000000.

With FFP-TF tuned and ODL2 properly adjusted, the output extinction ratio versus wavelength is plotted in Fig. 4. It can be noted that the output extinction ratio shows only a small fluctuation and is maintained above 10.3 dB from 1520 nm to 1560 nm except for a 4 nm gap around 1550 nm. The tuning gap is due to the configuration of the SFRL, because the laser wavelength cannot be the same as the input wavelength. The tuning gap is found to be wider than the FWHM bandwidth of the FFP-TF because the input pump power is much higher than the laser power in ring cavity so that farther wavelength spacing is needed to avoid the effect caused by pump beam. A FFP-TF with narrower FWHM bandwidth will increase the laser power in the ring cavity and improve the output performance. Moreover, the output extinction ratio is found to be lower around 1520 nm, which is caused by the asymmetry of the gain spectrum of the SOA.

 

Fig.4 . Output extinction ratio versus wavelength.

Download Full Size | PDF

B. Discussion: Key Factors in Output Performance

1. Pump Power

Since the logic operation is based on XGM of the SOA, the pump power is obviously one of the key factors in output performance. If we use the same input bit streams but adjust the tunable attenuator to reduce pump power 3 dBm at node a, the output waveform, as indicated in Fig. 5(a), shows almost a reversion compared with the pump signal [see Fig. 3(c)]. In this instance P₁₀ is no longer high enough to saturate the gain of SOA deeply, thus a submaximum of output optical power appears. Pump power should not be too low or too high so as to saturate the gain of SOA efficiently and also prevent the laser in the ring cavity from dying. In fact the experimental results shown in Fig. 3(d) are obtained when the input signal power is tuned to the optimum value. With the peak power of signal A and signal B tuned from 3.1 mw to 5.8 mw, output signal maintains the proper logic result. However, it should be noted that when the input signal power departs from the optimum value, the output performance is not as good as that shown in Fig. 3(d).

2. Total Length of Ring Cavity, or the Repetition Frequency of NRZ Signals

Like many other schemes that employ a fiber ring cavity, the total length of ring cavity, in other words, the repetition frequency of NRZ signals is another key factor in output performance. For general mode lock the repetition frequency of the pump signal requires tuning to be a precise harmonic of the fundamental frequency of the cavity. However, the requirement for frequency matching in the logic NOR operation is found to be different from that for mode lock. In this experiment the total length of the fiber ring cavity is about 20 m 4nm gap and the repetition frequency of the TFS is 5.007 GHz, so the actual bit rate of NRZ signals is 10.014 Gb/s in Fig. 3. Keeping the input bit streams and pump power unchanged but tuning the TFS up and down slightly, we change the output waveform. Fig. 5(b) is observed when the repetition frequency is tuned up 0.2 MHz (δf=0.2 MHz); the output waveform shows the characteristics of mode lock, such as narrower pulse width and higher peak power. Fig. 5(c) is observed when the repetition frequency is tuned down 0.5 MHz (δf=-0.5 MHz); the peak power of the two pulses is so different that the logic operation is not successful.

We note further that the small signal gain of the SOA and the fixed loss of the fiber ring cavity also influence the output performance directly.

We have not yet obtained the experimental results using input NRZ PRBS data 2³¹-1 because in this instance it will be more difficult to meet the requirement for frequency matching. Further researches are needed to testify the potential characteristics of this device.

 

Fig. 5. Illustration of output performance.

Download Full Size | PDF

(a) Pump power is attenuated 3 dBm, repetition frequency unchanged.

(b) Pump power unchanged, δf=0.2 MHz.

(c) Pump power unchanged, δf=-0.5 MHz.

4. Conclusion

A tunable, all-optical logic NOR gate based on XGM of a SOA is demonstrated at 10 Gb/s by employment of a fiber ring laser. It requires only a single SOA and two input signal beams without an additional light source, which makes this device quite simple and attractive. NOR operation with the output extinction ratio higher than 10.3dB is obtained experimentally over the 40 nm tuning range except for a 4 nm gap around the input wavelength. The output signal depends on the relative polarization of the two input logic signals. For similar polarizations of the input signals, interference effects may affect its performance. Input pump power and the length of the ring cavity (or the repetition frequency of the NRZ signals) are found to be key factors in output performance.