- Email: [email protected]
s all-optical logic NOR and OR gates using a semiconductor optical amplifier: Experimental demonstration and theoretical analysis
Available online at www.sciencedirect.com
Optics Communications 281 (2008) 1710–1715 www.elsevier.com/locate/optcom
40 Gb/s all-optical logic NOR and OR gates using a semiconductor optical amplifier: Experimental demonstration and theoretical analysis Jianji Dong, Xinliang Zhang *, Jing Xu, Dexiu Huang Wuhan National Laboratory for Optoelectronics, Huazhong University of Science and Technology, Wuhan 430074, China Received 17 March 2007; received in revised form 11 October 2007; accepted 12 November 2007
Abstract We experimentally and theoretically demonstrate 40 Gb/s all-optical logic NOR and OR gates based on a semiconductor optical amplifier (SOA) and a blue shifted optical bandpass filter (OBF). Two kinds of data formats are discussed, namely return-to-zero (RZ) format and nonreturn-to-zero (NRZ) format. The logic NOR and OR functions of RZ format are realized at the OBF detuning of 0.22 nm and 0.44 nm, respectively. The logic NOR function of NRZ format is realized at the OBF detuning of 0.24 nm. The simulation is in good agreement with the experimental results when the linewidth enhancement factor is 5.5. The simulation also shows that the SOA with large linewidth enhancement factor is preferred to achieve NOR and OR functions with good performance. The input data signal is of good pulsewidth-tolerance for NOR function, whereas not for OR function. The high Q factor could be obtained at narrow pulses injection. Ó 2007 Elsevier B.V. All rights reserved. Keywords: All-optical logic gate; Semiconductor optical amplifier (SOA); Cross phase modulation (XPM)
1. Introduction All-optical logic elements will be required to achieve high speed performance at data rates beyond 40 Gb/s because of the speed limitation set by the traditional electrical circuits [1]. Nonlinearities in semiconductor optical amplifier (SOA) have attracted considerable interest for realizing various logic gates. A variety of logic gates (including XOR, NOR, OR and HAND) based on cross phase modulation (XPM) have been presented [2] with advantages of high extinction ratio (ER) at the cost of complex interferometer configurations. Logic gates based on cross gain modulation (XGM) employing two-cascaded SOAs have been reported, such as AND logic and XOR logic with its application in half adders [3], AND logic
*
Corresponding author. Tel.: +86 27 87792367. E-mail address: [email protected] (X. Zhang).
0030-4018/$ - see front matter Ó 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.optcom.2007.11.054
and NOR logic realized by different power level control [4,5]. However the operation speed of XGM-based schemes is limited by the slow gain recovery time of the SOA. An effective solution to accelerate the gain recovery is with assistance of optical bandpass filter (OBF). Cho firstly presented the improved wavelength conversion (WC) scheme with a fiber Bragg grating filter [6,7]. Based on this approach high speed WC at 320 Gb/s has been demonstrated [8]. Moreover, multi-logic functions with AND, OR, and XOR gates at 10 Gb/s have been proposed with single SOA followed by a narrow OBF [9], and 40 Gb/s NOR function has been demonstrated as well [10]. However, data signals with ultrashort pulses (3 ps) are rigidly required to induce large phase shifts and chirp. In this paper, we experimentally and theoretically demonstrate 40 Gb/s logic NOR and OR gates based on a single SOA and a blue shifted OBF. The NOR function of return-to-zero (RZ) and nonreturn-to-zero (NRZ) formats is obtained with the XGM effect of the SOA. However a
J. Dong et al. / Optics Communications 281 (2008) 1710–1715
blue shifted OBF at a detuning of about 0.22 nm is necessary to accelerate the gain recovery since the gain recovery time is much longer than single bit duration. The OR function of RZ format is obtained with the transient XPM effect of the SOA. In this case, the blue shifted OBF at a detuning of 0.44 nm is used to select the blueshifted spectrum of the probe signal and convert the phase modulation to the intensity modulation. The Q factor of the NOR and OR functions dependent on the SOA a value and input pulsewidth is analyzed in numerical simulation. The simulation is in good agreement with the experimental results when a value is 5.5. The simulation shows that the NOR function has the potential advantages of pulsewidth-tolerance. 2. Operation principle and experimental setup The operation principle of the proposed logic functions is described in Fig. 1. Input signals with RZ and NRZ formats are illustrated in Fig. 1a and b, respectively. In Fig. 1a, two data signals with RZ format (Data A, Data B), combined with a continuous wave (CW) probe signal, are injected into the SOA. Due to XGM and XPM, the leading/trailing edge of the inverted probe signal is red/ blue shifted. Hence the probe spectrum is broadened. The subsequent OBF is used to reshape the probe spectrum at a central wavelength of kcw + Dkdet, which is different from the probe signal kcw. Our previous work showed that both inverted and non-inverted WC at 40 Gb/s could be obtained by adjusting the OBF detuning Dkdet properly [11]. Inspired by this work, we could obtain both OR logic and NOR functions with proper OBF detuning if two channels of the input data are launched into the SOA. When the OBF detuning is set to be properly large (negative), the OBF is used to reject the probe carrier and select the blue-shifted spectrum. Either Data A or Data B or both launched into the SOA will induce blue shifted spectrum, which fits in the OBF passband. If both Data A and Data B are absent, the OBF will block the probe carrier without blue shifted spectrum. Therefore the output is logic OR function. When the OBF detuning is small (negative), the XGM effect remains dominant in the SOA since the probe carrier is not suppressed. Hence either Data A or Data B or both launched into the SOA will induce polarity-inverted output, which is logic NOR function. The blue shifted OBF with small detuning is useful to accelerate the SOA
λcw
λcw + Δ λdet
a
λcw
1711
gain recovery. In our scheme, the SOA recovery time is much longer than the single bit period. If a single channel of input data can operate the XGM effect, the output will be the logic NOT of the input. In Fig. 1b, we replace Data A and Data B with NRZ format. The operation principle of NOR logic is similar to the case of RZ data signal injection. Note that the blue shifted OBF with small detuning is also useful to accelerate the SOA gain recovery even under NRZ signal injection. The output becomes logic NOT function of the input if a single NRZ data can operate the SOA. Unlike the case of RZ signal injection, one cannot observe the logic OR function of NRZ signal, because the NRZ amplitude keeps constant under consecutive bit ‘‘1”s injection, which could not induce wavelength shifts of the probe signal. The experimental setup for the proposed logic functions is shown in Fig. 2. The wavelengths of three CW beams generated by LD1, LD2, and LD3 are 1563.5 nm (kA), 1549.5 nm (kB), and 1557.32 nm (kcw), respectively. The data signals (kA and kB) are modulated by two Mach– Zehnder Modulators (MZMs) at 40 Gb/s to form 231-1 RZ pseudo random binary sequence (PRBS) signals. The duty cycle of these RZ pulses is 33%. The RZ format can be easily changed to NRZ format by removing the MZM driven by clock (CLK) signal. Then an erbium-doped fiber amplifier (EDFA1) and an attenuator (ATT) are used to adjust the RZ output peak power. Two data signals will be separated by the wavelength division multiplexer (WDM) and one of them is delayed for several bit periods by an optical delay line (ODL). Therefore, two data signals with different data pattern are obtained at the output of the optical coupler (OC2). The average power measured at the SOA input is 3.1 dBm (kA), 3.3 dBm (kB), 2.7 dBm (kcw), respectively. The SOA, manufactured by Kamelian Inc, is biased at 200 mA. No polarization devices are required because of the low polarization dependence (<0.5 dB). The SOA recovery time is about 90 ps, which is much longer than one bit period. A tunable OBF1 with 0.32 nm 3 dB-bandwidth has a detuning to the probe signal. Another EDFA2 and an OBF2 with 1 nm bandwidth are used to further amplify the converted signal power
λcw + Δλdet
b
Fig. 1. Schematic diagram of the logic functions based on a single SOA and a filter, (a) logic OR/NOR with RZ format, (b) logic NOR with NRZ format.
Fig. 2. Experimental setup of the proposed logic functions, BPG: bit pattern generator; MZM: Mach–Zehnder Modulators; ATT: attenuator; OC: optical coupler; ODL: optical delay line; OBF: optical bandpass filter; OSA: optical spectrum analyzer; CSA: communication signal analyzer.
1712
J. Dong et al. / Optics Communications 281 (2008) 1710–1715
and eliminate the crosstalk from the data signals. Finally, the optical spectrum analyzer (OSA) with Anritsu MS9710C series is used to observe the output optical spectrum and the communication signal analyzer (CSA) is used to measure the waveform of the converted signal. The CSA is Tektronix CSA8000B series with the bandwidth of 65 GHz, which can automatically measure the ER and Q factor of the captured waveforms. 3. Results and discussion In order to investigate the relationship between the output ER and the input data wavelength, we switch off LD2 firstly and validate the NOT function of Data A. When OBF1 detuning is optimized at 0.15 nm and 0.24 nm, we obtain the NOT logic waveforms of RZ and NRZ formats, respectively. Fig. 3 shows Data A with RZ format (i) and its NOT logic waveform (ii), Data A with NRZ format (iii) and the NOT logic waveform (iv). We can observe correct logic operation results for all input data stream. The measured output ER of the NOT function is 8.1 dB for NRZ format and 7.3 dB for RZ format. By changing the wavelength of tunable LD1, we investigate the output ER under the optimal D kdet value with respect to the wavelength of Data A, as shown in Fig. 4. The solid line and dash line represent RZ and NRZ NOT functions, respectively. The output ER fluctuates around 7 dB from 1540 nm to 1563 nm, but is sharply degraded at short wavelength (below 1540 nm). The fluctuation of output ER results from the wavelength-dependent gain coefficient. In addition, Fig. 4 suggests that the data signals with longer wavelength are preferred to obtain higher output ER for both RZ and NRZ signals. Now, we turn on LD2 so that both Data A and Data B operate normally. Fig. 5 shows the experimental results of NOR function (iii) and OR function (iv) when OBF1 is blue shifted by 0.22 nm and 0.44 nm, respectively. The input waveforms of Data A and Data B are shown in Fig. 5i and ii. As for NOR function, the OBF1 detuning is as small as 0.22 nm. The eye diagram of NOR logic is clear and open except some noise on both level ‘‘1” and level ‘‘0”. The Q factor is 6.5 for NOR logic function. As for OR function, the OBF1 detuning is as large as
Fig. 3. Waveforms of input Data A with RZ format (i) and NRZ format (iii), waveforms of output NOT with RZ format (ii) and NRZ format (iv).
8
6
NRZ NOT RZ NOT
4
2 1530
1535
1540
1545
1550
1555
1560
Fig. 4. Output ER as a function of the data wavelength in NOT function. Solid line: RZ format; Dash line: NRZ format.
Fig. 5. Experimental results for NOR and OR functions with RZ format, (i) waveform of Data A, (ii) waveform of Data B, (iii) NOR output waveform, (iv) OR output waveform.
0.44 nm. Similarly, we measure that the Q factor is 6.2. However, we can see that some pattern effect and amplitude noise exist. An extended SOA rate equation model is used to simulate the logic functions at 40 Gb/s. The model includes ultrafast carrier dynamics with intraband effect, such as spectral hole burning (SHB) and carrier heating (CH). A detailed SOA model description and parameter lists were presented in Ref. [12]. The linewidth enhancement factor a is set as 5.5 first. In the simulation, the initial parameters are set same to the experimental settings. The simulated results for logic NOR and OR functions with RZ format are shown in Fig. 6iii and v. The calculated Q factor is 6.5 for the NOR function at a detuning of 0.22 nm and 6.1 for the OR function at a detuning of 0.44 nm. The Q factor is calculated according to Ref. [13], which is 0 defined by Q ¼ Pr11 P , where P1 and P0 are the average þr0 power of ‘‘1” and ‘‘0” signal, respectively, while r1 andr0 are the standard deviation of ‘‘1” and ‘‘0” signal. The simulation shows serious pattern effects and amplitude noise, and conforms to the experimental results. As mentioned above, the logic NOR and OR are realized at the OBF1 detuning of 0.22 nm and 0.44 nm, respectively. When a value varies from 3 to 7, we simulate these two logic functions and calculate the Q factor, as
J. Dong et al. / Optics Communications 281 (2008) 1710–1715
Fig. 6. Simulation results for NOR and OR functions with RZ format, (i) waveform of Data A, (ii) waveform of Data B, (iii) NOR output waveform with CW injection, (iv) NOR output waveform with pulsed clock injection, (v) OR output waveform with CW injection.
1713
function, which is shown in Fig. 6iv. The peak power of the clock signal and the data signal are 0.2 mW and 10 mW, respectively, and the OBF detuning is 0.32 nm. We can see that the output ER is as low as 5.6 dB and the power ‘‘0”s are not sufficiently suppressed. The format-maintaining logic function is advantageous in cascaded logic schemes. However, the pulsed clock source will degrade the logic operation. Now we set the input data to NRZ format. When OBF1 detuning is 0.24 nm, the output waveform of NOR function is shown in Fig. 8(iv). The waveforms of Data A, Data B, combination of Data A and Data B are shown in Fig. 8i–iii, respectively. We observe that correct logic is obtained for all bit sequences. The measured Q factor is 4.1. However, some ‘‘ghost” pulses appear at level ‘‘0” of NOR output, which could degrade the output performance of logic operation to some extent. Form Fig. 8iii, we notice that there are leading or trailing edges at the transition of P10 resulting from a single input ‘‘one” and P11 resulting from two input ‘‘one”s, which lead to the phase modulation of probe carrier. The phase modulation will be partially converted to intensity modulation by the blue shifted OBF1, hence the ‘‘ghost” pulses appear. An optional solution to avoid the ghost pulses is reducing the OBF1 detuning but the pattern effects will be introduced as the penalty. Therefore, there is a tradeoff between weak pattern effects and ghost pulses affected by OBF1 detuning. Fig. 9 shows the simulated results of NOR function with NRZ format at detuning of 0.24 nm when a value is 5.5. The simulated Q factor is 4.0, which shows good agreements with the experimental results. In fact, NRZ format can be regarded as a special RZ format with 100% duty cycle. Since logic NOR can be obtained with both RZ format and NRZ format, the input data signal is of good pulsewidth-tolerance. While logic OR is obtained with only RZ format. The XPM in the SOA plays important role in NOR and OR functions, therefore we are concerned about the impact of a on the pulsewidth-tolerance. Fig. 10 shows the simulated Q factor of NOR and OR functions as a function of the input pulsewdith when a is 6. The Q factor decreases as the
Fig. 7. Q factor for NOR and OR functions as a function of the a value.
shown in Fig. 7. We find the calculated Q factor has a good agreement with the experimental results when a value is 5.5, so we estimate a value of the SOA at about 5.5. The insets of Fig. 7 are eye diagrams of the logic NOR and OR. Furthermore, Fig. 7 shows that good performance of the logic functions can be obtained when using the SOA with large a value. It should be noted that logic NOR function shown in Fig. 6iii has an output with inverted RZ format, which is different from the input format, however, the bit error decision is not affected. In fact, the NOR function can maintain the non-inverted format if a synchronized clock replaces the CW light. We simulate the format-maintaining NOR
Fig. 8. Experimental results for NOR function with NRZ format. (i) waveform of Data A, (ii) waveform of Data B, (iii) combined waveform of Data A and Data B, (iv) NOR output waveform.
1714
J. Dong et al. / Optics Communications 281 (2008) 1710–1715
Fig. 9. Simulation results for NOR function with NRZ format. (i) waveform of Data A, (ii) waveform of Data B, (iii) combined waveform of Data A and Data B, (iv) NOR output waveform.
Fig. 11. Q factor as a function of the input pulsewidth when a is 3.
value is preferred to achieve NOR and OR functions with good performance, because the XPM effect is significant in both NOR and OR functions. In the OR function, the XPM induces shifted spectrum, which fits in the filter passband. In the NOR function, the XPM induces frequency chirp, which can compensate the slow recovery of SOA with the filter’s assistance. Although a wide range of input pulsewidth is tolerable in the NOR function, narrow pulses are preferred in order to get high Q factor, which is also suitable for the OR function. 4. Conclusion
Fig. 10. Q factor as a function of the input pulsewidth when a is 6.
pulsewidth increases for both NOR and OR functions. As for OR function, the Q factor decreases sharply. Good eye pattern is shown when the input pulsewidth is 4 ps, however the output pulse is broadened at 10 ps input pulsewidth and the Q factor drops to 4.5. We cannot observe OR function at 25 ps input pulsewidth (NRZ format). As for NOR function, good eye pattern is shown at 4 ps input pulsewidth and eye pattern with some amplitude noise is observed at 25 ps input pulsewidth, although the Q factor decreases as input pulsewidth increases. When a is 3, we investigate the impact of the input pulsewidth on Q factor, as shown in Fig. 11. We notice that Q factor when a = 3 is much lower than that when a = 6 at identical pulsewidth. Especially when the pulsewidth is 25 ps, the eye pattern of NOR function shows poor quality with large amplitude noise and small opening. When the pulsewidth is 10 ps, the eye pattern of OR function is poor and the output pulses are broadened as NRZ shape. From Figs. 10 and 11, we can infer that the SOA with large a
We have demonstrated 40 Gb/s all-optical logic OR and NOR functions based on a single SOA and a blue-shifted OBF. The logic NOR and OR functions of RZ signal are realized at the OBF detuning of 0.22 nm and 0.44 nm, respectively. The logic NOR function of NRZ signal is realized at the OBF detuning of 0.24 nm. Experimental results show that the data signals with longer wavelength (over 1540 nm) are preferred to obtain higher output ER. The simulations are in good agreement with the experimental results when a is 5.5. The simulations also show that the SOA with large a value is preferred to achieve NOR and OR functions with good performance. The input signal is good pulsewidth-tolerant for NOR function, but not for OR function. The high Q factor could be obtained at narrow pulses injection. Acknowledgements This work was partially supported by the National High Technology Developing Program of China (Grant No. 2006AA03Z0414), the National Natural Science Foundation of China (Grant No. 60407001), the Science Fund for Distinguished Young Scholars of Hubei Province (Grant No. 2006ABB017).
J. Dong et al. / Optics Communications 281 (2008) 1710–1715
References [1] H. Dong, H. Sun, Q. Wang, N.K. Dutta, J. Jaques, Opt. Commun. 265 (2006) 79. [2] J.-Y. Kim, J.-M. Kang, T.-Y. Kim, S.-K. Han, Electron. Lett. 42 (2006) 303. [3] S.-H. Kim, J.-H. Kim, J.-W. Choi, C.-W. Son, Y.-T. Byun, et al., Opt. Express 14 (2006) 10693. [4] X. Zhang, Y. Wang, J. Sun, D. Liu, D. Huang, Opt. Express 12 (2004) 361. [5] A. Hamie, A. Sharaiha, M. Guegan, B. Pucel, IEEE Photon. Technol. Lett. 14 (2002) 1439. [6] P.S. Cho, D. Mahgerefteh, J. Goldhar, G.L. Burdge, IEEE Photon. Technol. Lett. 10 (1998) 66. [7] P.S. Cho, D. Mahgerefteh, J. Goldhar, L.G. Joneckis, G.L. Burdge, Electron. Lett. 34 (1998) 371.
1715
[8] Y. Liu, E. Tangdiongga, Z. Li, H. deWaardt, A.M.J. Koonen, Errorfree 320 Gb/s SOA-based wavelength conversion using optical filtering, OFC 2006 PDP28, 2006. [9] Z. Li, Y. Liu, S. Zhang, H. Ju, H. de Waardt, G.D. Khoe, H.J.S. Dorren, D. Lenstra, Electron. Lett. 41 (2005) 1397. [10] J. Zhang, J. Wu, C. Feng, G. Zhou, K. Xu, J. Lin, Opt. Eng. 45 (080503) (2006). [11] J. Dong, X. Zhang, J. Xu, D. Huang, S. Fu, P. Shum, L. Zhang, Y. Gong, 40 Gb/s both inverted and non-inverted wavelength conversion based on transient XPM of SOA, Anaheim, USA, 2007, OFC. [12] J. Dong, S. Fu, X. Zhang, P. Shum, J. Xu, D. Huang, Opt. Commun. 270 (2007) 238. [13] Q. Wang, H. Dong, G. Zhu, H. Sun, J. Jaques, A.B. Piccirilli, N.K. Dutta, Opt. Commun. 260 (2006) 81.
Optics Communications 281 (2008) 1710–1715 www.elsevier.com/locate/optcom
40 Gb/s all-optical logic NOR and OR gates using a semiconductor optical amplifier: Experimental demonstration and theoretical analysis Jianji Dong, Xinliang Zhang *, Jing Xu, Dexiu Huang Wuhan National Laboratory for Optoelectronics, Huazhong University of Science and Technology, Wuhan 430074, China Received 17 March 2007; received in revised form 11 October 2007; accepted 12 November 2007
Abstract We experimentally and theoretically demonstrate 40 Gb/s all-optical logic NOR and OR gates based on a semiconductor optical amplifier (SOA) and a blue shifted optical bandpass filter (OBF). Two kinds of data formats are discussed, namely return-to-zero (RZ) format and nonreturn-to-zero (NRZ) format. The logic NOR and OR functions of RZ format are realized at the OBF detuning of 0.22 nm and 0.44 nm, respectively. The logic NOR function of NRZ format is realized at the OBF detuning of 0.24 nm. The simulation is in good agreement with the experimental results when the linewidth enhancement factor is 5.5. The simulation also shows that the SOA with large linewidth enhancement factor is preferred to achieve NOR and OR functions with good performance. The input data signal is of good pulsewidth-tolerance for NOR function, whereas not for OR function. The high Q factor could be obtained at narrow pulses injection. Ó 2007 Elsevier B.V. All rights reserved. Keywords: All-optical logic gate; Semiconductor optical amplifier (SOA); Cross phase modulation (XPM)
1. Introduction All-optical logic elements will be required to achieve high speed performance at data rates beyond 40 Gb/s because of the speed limitation set by the traditional electrical circuits [1]. Nonlinearities in semiconductor optical amplifier (SOA) have attracted considerable interest for realizing various logic gates. A variety of logic gates (including XOR, NOR, OR and HAND) based on cross phase modulation (XPM) have been presented [2] with advantages of high extinction ratio (ER) at the cost of complex interferometer configurations. Logic gates based on cross gain modulation (XGM) employing two-cascaded SOAs have been reported, such as AND logic and XOR logic with its application in half adders [3], AND logic
*
Corresponding author. Tel.: +86 27 87792367. E-mail address: [email protected] (X. Zhang).
0030-4018/$ - see front matter Ó 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.optcom.2007.11.054
and NOR logic realized by different power level control [4,5]. However the operation speed of XGM-based schemes is limited by the slow gain recovery time of the SOA. An effective solution to accelerate the gain recovery is with assistance of optical bandpass filter (OBF). Cho firstly presented the improved wavelength conversion (WC) scheme with a fiber Bragg grating filter [6,7]. Based on this approach high speed WC at 320 Gb/s has been demonstrated [8]. Moreover, multi-logic functions with AND, OR, and XOR gates at 10 Gb/s have been proposed with single SOA followed by a narrow OBF [9], and 40 Gb/s NOR function has been demonstrated as well [10]. However, data signals with ultrashort pulses (3 ps) are rigidly required to induce large phase shifts and chirp. In this paper, we experimentally and theoretically demonstrate 40 Gb/s logic NOR and OR gates based on a single SOA and a blue shifted OBF. The NOR function of return-to-zero (RZ) and nonreturn-to-zero (NRZ) formats is obtained with the XGM effect of the SOA. However a
J. Dong et al. / Optics Communications 281 (2008) 1710–1715
blue shifted OBF at a detuning of about 0.22 nm is necessary to accelerate the gain recovery since the gain recovery time is much longer than single bit duration. The OR function of RZ format is obtained with the transient XPM effect of the SOA. In this case, the blue shifted OBF at a detuning of 0.44 nm is used to select the blueshifted spectrum of the probe signal and convert the phase modulation to the intensity modulation. The Q factor of the NOR and OR functions dependent on the SOA a value and input pulsewidth is analyzed in numerical simulation. The simulation is in good agreement with the experimental results when a value is 5.5. The simulation shows that the NOR function has the potential advantages of pulsewidth-tolerance. 2. Operation principle and experimental setup The operation principle of the proposed logic functions is described in Fig. 1. Input signals with RZ and NRZ formats are illustrated in Fig. 1a and b, respectively. In Fig. 1a, two data signals with RZ format (Data A, Data B), combined with a continuous wave (CW) probe signal, are injected into the SOA. Due to XGM and XPM, the leading/trailing edge of the inverted probe signal is red/ blue shifted. Hence the probe spectrum is broadened. The subsequent OBF is used to reshape the probe spectrum at a central wavelength of kcw + Dkdet, which is different from the probe signal kcw. Our previous work showed that both inverted and non-inverted WC at 40 Gb/s could be obtained by adjusting the OBF detuning Dkdet properly [11]. Inspired by this work, we could obtain both OR logic and NOR functions with proper OBF detuning if two channels of the input data are launched into the SOA. When the OBF detuning is set to be properly large (negative), the OBF is used to reject the probe carrier and select the blue-shifted spectrum. Either Data A or Data B or both launched into the SOA will induce blue shifted spectrum, which fits in the OBF passband. If both Data A and Data B are absent, the OBF will block the probe carrier without blue shifted spectrum. Therefore the output is logic OR function. When the OBF detuning is small (negative), the XGM effect remains dominant in the SOA since the probe carrier is not suppressed. Hence either Data A or Data B or both launched into the SOA will induce polarity-inverted output, which is logic NOR function. The blue shifted OBF with small detuning is useful to accelerate the SOA
λcw
λcw + Δ λdet
a
λcw
1711
gain recovery. In our scheme, the SOA recovery time is much longer than the single bit period. If a single channel of input data can operate the XGM effect, the output will be the logic NOT of the input. In Fig. 1b, we replace Data A and Data B with NRZ format. The operation principle of NOR logic is similar to the case of RZ data signal injection. Note that the blue shifted OBF with small detuning is also useful to accelerate the SOA gain recovery even under NRZ signal injection. The output becomes logic NOT function of the input if a single NRZ data can operate the SOA. Unlike the case of RZ signal injection, one cannot observe the logic OR function of NRZ signal, because the NRZ amplitude keeps constant under consecutive bit ‘‘1”s injection, which could not induce wavelength shifts of the probe signal. The experimental setup for the proposed logic functions is shown in Fig. 2. The wavelengths of three CW beams generated by LD1, LD2, and LD3 are 1563.5 nm (kA), 1549.5 nm (kB), and 1557.32 nm (kcw), respectively. The data signals (kA and kB) are modulated by two Mach– Zehnder Modulators (MZMs) at 40 Gb/s to form 231-1 RZ pseudo random binary sequence (PRBS) signals. The duty cycle of these RZ pulses is 33%. The RZ format can be easily changed to NRZ format by removing the MZM driven by clock (CLK) signal. Then an erbium-doped fiber amplifier (EDFA1) and an attenuator (ATT) are used to adjust the RZ output peak power. Two data signals will be separated by the wavelength division multiplexer (WDM) and one of them is delayed for several bit periods by an optical delay line (ODL). Therefore, two data signals with different data pattern are obtained at the output of the optical coupler (OC2). The average power measured at the SOA input is 3.1 dBm (kA), 3.3 dBm (kB), 2.7 dBm (kcw), respectively. The SOA, manufactured by Kamelian Inc, is biased at 200 mA. No polarization devices are required because of the low polarization dependence (<0.5 dB). The SOA recovery time is about 90 ps, which is much longer than one bit period. A tunable OBF1 with 0.32 nm 3 dB-bandwidth has a detuning to the probe signal. Another EDFA2 and an OBF2 with 1 nm bandwidth are used to further amplify the converted signal power
λcw + Δλdet
b
Fig. 1. Schematic diagram of the logic functions based on a single SOA and a filter, (a) logic OR/NOR with RZ format, (b) logic NOR with NRZ format.
Fig. 2. Experimental setup of the proposed logic functions, BPG: bit pattern generator; MZM: Mach–Zehnder Modulators; ATT: attenuator; OC: optical coupler; ODL: optical delay line; OBF: optical bandpass filter; OSA: optical spectrum analyzer; CSA: communication signal analyzer.
1712
J. Dong et al. / Optics Communications 281 (2008) 1710–1715
and eliminate the crosstalk from the data signals. Finally, the optical spectrum analyzer (OSA) with Anritsu MS9710C series is used to observe the output optical spectrum and the communication signal analyzer (CSA) is used to measure the waveform of the converted signal. The CSA is Tektronix CSA8000B series with the bandwidth of 65 GHz, which can automatically measure the ER and Q factor of the captured waveforms. 3. Results and discussion In order to investigate the relationship between the output ER and the input data wavelength, we switch off LD2 firstly and validate the NOT function of Data A. When OBF1 detuning is optimized at 0.15 nm and 0.24 nm, we obtain the NOT logic waveforms of RZ and NRZ formats, respectively. Fig. 3 shows Data A with RZ format (i) and its NOT logic waveform (ii), Data A with NRZ format (iii) and the NOT logic waveform (iv). We can observe correct logic operation results for all input data stream. The measured output ER of the NOT function is 8.1 dB for NRZ format and 7.3 dB for RZ format. By changing the wavelength of tunable LD1, we investigate the output ER under the optimal D kdet value with respect to the wavelength of Data A, as shown in Fig. 4. The solid line and dash line represent RZ and NRZ NOT functions, respectively. The output ER fluctuates around 7 dB from 1540 nm to 1563 nm, but is sharply degraded at short wavelength (below 1540 nm). The fluctuation of output ER results from the wavelength-dependent gain coefficient. In addition, Fig. 4 suggests that the data signals with longer wavelength are preferred to obtain higher output ER for both RZ and NRZ signals. Now, we turn on LD2 so that both Data A and Data B operate normally. Fig. 5 shows the experimental results of NOR function (iii) and OR function (iv) when OBF1 is blue shifted by 0.22 nm and 0.44 nm, respectively. The input waveforms of Data A and Data B are shown in Fig. 5i and ii. As for NOR function, the OBF1 detuning is as small as 0.22 nm. The eye diagram of NOR logic is clear and open except some noise on both level ‘‘1” and level ‘‘0”. The Q factor is 6.5 for NOR logic function. As for OR function, the OBF1 detuning is as large as
Fig. 3. Waveforms of input Data A with RZ format (i) and NRZ format (iii), waveforms of output NOT with RZ format (ii) and NRZ format (iv).
8
6
NRZ NOT RZ NOT
4
2 1530
1535
1540
1545
1550
1555
1560
Fig. 4. Output ER as a function of the data wavelength in NOT function. Solid line: RZ format; Dash line: NRZ format.
Fig. 5. Experimental results for NOR and OR functions with RZ format, (i) waveform of Data A, (ii) waveform of Data B, (iii) NOR output waveform, (iv) OR output waveform.
0.44 nm. Similarly, we measure that the Q factor is 6.2. However, we can see that some pattern effect and amplitude noise exist. An extended SOA rate equation model is used to simulate the logic functions at 40 Gb/s. The model includes ultrafast carrier dynamics with intraband effect, such as spectral hole burning (SHB) and carrier heating (CH). A detailed SOA model description and parameter lists were presented in Ref. [12]. The linewidth enhancement factor a is set as 5.5 first. In the simulation, the initial parameters are set same to the experimental settings. The simulated results for logic NOR and OR functions with RZ format are shown in Fig. 6iii and v. The calculated Q factor is 6.5 for the NOR function at a detuning of 0.22 nm and 6.1 for the OR function at a detuning of 0.44 nm. The Q factor is calculated according to Ref. [13], which is 0 defined by Q ¼ Pr11 P , where P1 and P0 are the average þr0 power of ‘‘1” and ‘‘0” signal, respectively, while r1 andr0 are the standard deviation of ‘‘1” and ‘‘0” signal. The simulation shows serious pattern effects and amplitude noise, and conforms to the experimental results. As mentioned above, the logic NOR and OR are realized at the OBF1 detuning of 0.22 nm and 0.44 nm, respectively. When a value varies from 3 to 7, we simulate these two logic functions and calculate the Q factor, as
J. Dong et al. / Optics Communications 281 (2008) 1710–1715
Fig. 6. Simulation results for NOR and OR functions with RZ format, (i) waveform of Data A, (ii) waveform of Data B, (iii) NOR output waveform with CW injection, (iv) NOR output waveform with pulsed clock injection, (v) OR output waveform with CW injection.
1713
function, which is shown in Fig. 6iv. The peak power of the clock signal and the data signal are 0.2 mW and 10 mW, respectively, and the OBF detuning is 0.32 nm. We can see that the output ER is as low as 5.6 dB and the power ‘‘0”s are not sufficiently suppressed. The format-maintaining logic function is advantageous in cascaded logic schemes. However, the pulsed clock source will degrade the logic operation. Now we set the input data to NRZ format. When OBF1 detuning is 0.24 nm, the output waveform of NOR function is shown in Fig. 8(iv). The waveforms of Data A, Data B, combination of Data A and Data B are shown in Fig. 8i–iii, respectively. We observe that correct logic is obtained for all bit sequences. The measured Q factor is 4.1. However, some ‘‘ghost” pulses appear at level ‘‘0” of NOR output, which could degrade the output performance of logic operation to some extent. Form Fig. 8iii, we notice that there are leading or trailing edges at the transition of P10 resulting from a single input ‘‘one” and P11 resulting from two input ‘‘one”s, which lead to the phase modulation of probe carrier. The phase modulation will be partially converted to intensity modulation by the blue shifted OBF1, hence the ‘‘ghost” pulses appear. An optional solution to avoid the ghost pulses is reducing the OBF1 detuning but the pattern effects will be introduced as the penalty. Therefore, there is a tradeoff between weak pattern effects and ghost pulses affected by OBF1 detuning. Fig. 9 shows the simulated results of NOR function with NRZ format at detuning of 0.24 nm when a value is 5.5. The simulated Q factor is 4.0, which shows good agreements with the experimental results. In fact, NRZ format can be regarded as a special RZ format with 100% duty cycle. Since logic NOR can be obtained with both RZ format and NRZ format, the input data signal is of good pulsewidth-tolerance. While logic OR is obtained with only RZ format. The XPM in the SOA plays important role in NOR and OR functions, therefore we are concerned about the impact of a on the pulsewidth-tolerance. Fig. 10 shows the simulated Q factor of NOR and OR functions as a function of the input pulsewdith when a is 6. The Q factor decreases as the
Fig. 7. Q factor for NOR and OR functions as a function of the a value.
shown in Fig. 7. We find the calculated Q factor has a good agreement with the experimental results when a value is 5.5, so we estimate a value of the SOA at about 5.5. The insets of Fig. 7 are eye diagrams of the logic NOR and OR. Furthermore, Fig. 7 shows that good performance of the logic functions can be obtained when using the SOA with large a value. It should be noted that logic NOR function shown in Fig. 6iii has an output with inverted RZ format, which is different from the input format, however, the bit error decision is not affected. In fact, the NOR function can maintain the non-inverted format if a synchronized clock replaces the CW light. We simulate the format-maintaining NOR
Fig. 8. Experimental results for NOR function with NRZ format. (i) waveform of Data A, (ii) waveform of Data B, (iii) combined waveform of Data A and Data B, (iv) NOR output waveform.
1714
J. Dong et al. / Optics Communications 281 (2008) 1710–1715
Fig. 9. Simulation results for NOR function with NRZ format. (i) waveform of Data A, (ii) waveform of Data B, (iii) combined waveform of Data A and Data B, (iv) NOR output waveform.
Fig. 11. Q factor as a function of the input pulsewidth when a is 3.
value is preferred to achieve NOR and OR functions with good performance, because the XPM effect is significant in both NOR and OR functions. In the OR function, the XPM induces shifted spectrum, which fits in the filter passband. In the NOR function, the XPM induces frequency chirp, which can compensate the slow recovery of SOA with the filter’s assistance. Although a wide range of input pulsewidth is tolerable in the NOR function, narrow pulses are preferred in order to get high Q factor, which is also suitable for the OR function. 4. Conclusion
Fig. 10. Q factor as a function of the input pulsewidth when a is 6.
pulsewidth increases for both NOR and OR functions. As for OR function, the Q factor decreases sharply. Good eye pattern is shown when the input pulsewidth is 4 ps, however the output pulse is broadened at 10 ps input pulsewidth and the Q factor drops to 4.5. We cannot observe OR function at 25 ps input pulsewidth (NRZ format). As for NOR function, good eye pattern is shown at 4 ps input pulsewidth and eye pattern with some amplitude noise is observed at 25 ps input pulsewidth, although the Q factor decreases as input pulsewidth increases. When a is 3, we investigate the impact of the input pulsewidth on Q factor, as shown in Fig. 11. We notice that Q factor when a = 3 is much lower than that when a = 6 at identical pulsewidth. Especially when the pulsewidth is 25 ps, the eye pattern of NOR function shows poor quality with large amplitude noise and small opening. When the pulsewidth is 10 ps, the eye pattern of OR function is poor and the output pulses are broadened as NRZ shape. From Figs. 10 and 11, we can infer that the SOA with large a
We have demonstrated 40 Gb/s all-optical logic OR and NOR functions based on a single SOA and a blue-shifted OBF. The logic NOR and OR functions of RZ signal are realized at the OBF detuning of 0.22 nm and 0.44 nm, respectively. The logic NOR function of NRZ signal is realized at the OBF detuning of 0.24 nm. Experimental results show that the data signals with longer wavelength (over 1540 nm) are preferred to obtain higher output ER. The simulations are in good agreement with the experimental results when a is 5.5. The simulations also show that the SOA with large a value is preferred to achieve NOR and OR functions with good performance. The input signal is good pulsewidth-tolerant for NOR function, but not for OR function. The high Q factor could be obtained at narrow pulses injection. Acknowledgements This work was partially supported by the National High Technology Developing Program of China (Grant No. 2006AA03Z0414), the National Natural Science Foundation of China (Grant No. 60407001), the Science Fund for Distinguished Young Scholars of Hubei Province (Grant No. 2006ABB017).
J. Dong et al. / Optics Communications 281 (2008) 1710–1715
References [1] H. Dong, H. Sun, Q. Wang, N.K. Dutta, J. Jaques, Opt. Commun. 265 (2006) 79. [2] J.-Y. Kim, J.-M. Kang, T.-Y. Kim, S.-K. Han, Electron. Lett. 42 (2006) 303. [3] S.-H. Kim, J.-H. Kim, J.-W. Choi, C.-W. Son, Y.-T. Byun, et al., Opt. Express 14 (2006) 10693. [4] X. Zhang, Y. Wang, J. Sun, D. Liu, D. Huang, Opt. Express 12 (2004) 361. [5] A. Hamie, A. Sharaiha, M. Guegan, B. Pucel, IEEE Photon. Technol. Lett. 14 (2002) 1439. [6] P.S. Cho, D. Mahgerefteh, J. Goldhar, G.L. Burdge, IEEE Photon. Technol. Lett. 10 (1998) 66. [7] P.S. Cho, D. Mahgerefteh, J. Goldhar, L.G. Joneckis, G.L. Burdge, Electron. Lett. 34 (1998) 371.
1715
[8] Y. Liu, E. Tangdiongga, Z. Li, H. deWaardt, A.M.J. Koonen, Errorfree 320 Gb/s SOA-based wavelength conversion using optical filtering, OFC 2006 PDP28, 2006. [9] Z. Li, Y. Liu, S. Zhang, H. Ju, H. de Waardt, G.D. Khoe, H.J.S. Dorren, D. Lenstra, Electron. Lett. 41 (2005) 1397. [10] J. Zhang, J. Wu, C. Feng, G. Zhou, K. Xu, J. Lin, Opt. Eng. 45 (080503) (2006). [11] J. Dong, X. Zhang, J. Xu, D. Huang, S. Fu, P. Shum, L. Zhang, Y. Gong, 40 Gb/s both inverted and non-inverted wavelength conversion based on transient XPM of SOA, Anaheim, USA, 2007, OFC. [12] J. Dong, S. Fu, X. Zhang, P. Shum, J. Xu, D. Huang, Opt. Commun. 270 (2007) 238. [13] Q. Wang, H. Dong, G. Zhu, H. Sun, J. Jaques, A.B. Piccirilli, N.K. Dutta, Opt. Commun. 260 (2006) 81.