Open Access
Add to BibSonomyAbstract
We propose and experimentally demonstrate a hybrid WDM/OCDMA ring with a dynamic add/drop function based on Fourier code for local area networks. Dynamic function is implemented by mechanically tuning the Fourier encoder/decoder for optical code division multiple access (OCDMA) encoding/decoding. Wavelength division multiplexing (WDM) is utilized for node assignment and 4-chip Fourier code recovers the matched signal from the codes. For an optical source well adapted to WDM channels and its short optical pulse generation, reflective semiconductor optical amplifiers (RSOAs) are used with a fiber Bragg grating (FBG) and gain-switched. To demonstrate we experimentally investigated a two-node hybrid WDM/OCDMA ring with a 4-chip Fourier encoder/decoder fabricated by cascading four FBGs with the bit error rate (BER) of <10−9 for the node span of 10.64 km at 1.25 Gb/s.
©2011 Optical Society of America
1. Introduction
In local area networks (LANs) that include a large number of users, bandwidth as well as optical power distribution to each user is an important issue. Recently, an increase in the bandwidth occupied by services necessitates that the transmission capacity of LANs be upgraded to support such services. Time division multiplexing (TDM), wavelength division multiplexing (WDM), and optical code division multiplexing (OCDM) can be used alone or in their combined form [1–5] as a part of increasing bandwidth in LANs. To increase transmission capacity, the WDM could be combined with TDM or OCDM because it has to avoid collisions with its own wavelength. By introducing hybrid WDM/OCDM, the number of required wavelength could be drastically reduced and node configuration could be very simple. Also, the WDM/OCDMA configuration provides asynchronous access to the ring by OCDM coding the synchronous optical network (SONET) or Ethernet signal to be added to the ring without data collision. On the other hand, in WDM-SONET ring, TDM access protocol for the upstream signal should be used. Therefore, the hybrid form of WDM and OCDM could be an especially good candidate for LANs requiring high bandwidth with security and easy accessibility.
In a ring configuration consisting of WDM and OCDM, system performance depends on wavelength channel cross-talk and code interference. The former is due to imperfect channel filtering in WDM, and the latter is caused by cross-correlation between the neighboring optical coded channels. In optical code division multiple access (OCDMA), the code interference depends on the cross-correlation between the orthogonal codes of users. To decrease the interference, a few orthogonal codes such as binary phase shift keying (BPSK) using a photonic crystal [6] and quarternary phase shift keying (QPSK) using a super-structured fiber Bragg grating (SSFBG) [7] have been used. The QPSK orthogonal code has been known to provide a greatly improved cross-correlation characteristic compared with the BPSK orthogonal code [8]. Furthermore, the Fourier code, as one of the QPSK orthogonal codes, inherits the lowered code interference from the orthogonality [9,10]. Furthermore, the encoder/decoder based on the Fourier code has an advantage of easy code adaptation because the Fourier code has a same phase difference among the orthogonal codes. This characteristic of the encoder/decoder is very useful in a ring configuration with the requirement of dynamically adding or dropping coded channels on the operating wavelengths [11,12].
In this paper, we propose and experimentally demonstrate a two-node hybrid WDM/OCDMA ring with a dynamic add/drop function based on the Fourier code for LAN applications. The Fourier encoder/decoder is fabricated by cascading four fiber Bragg gratings (FBGs). Different from the conventional methods using a wavelength-managed mode lock laser diode (MLLD) to generate a short pulse train [4,5], a reflective semiconductor optical amplifier (RSOA) is used with a wavelength-selective reflector to provide a WDM capability and is gain-switched to generate a short pulse train. This short pulse train is coupled to another RSOA and is on-off modulated by the modulation signal of 1.25 Gb/s.
2. Proposed WDM/OCDMA ring configuration
Figure 1 shows the proposed WDM/OCDMA on a single fiber ring which is composed of multiple nodes. And the lower part of Fig. 1 shows the signal flows between the nodes and their block diagram. At each node, one wavelength is assumed to be added and dropped, and N different Fourier codes from the N-chip Fourier encoder/decoder can be generated. Among the wavelengths guided along the ring, the wavelength assigned to the node is dropped, and the same wavelength or different wavelength with new information can be added to the ring through the node. To generate a Fourier coded optical signal, an optical source with a specific wavelength is necessary. Here, we use an RSOA which has broadband spectrum but its lasing wavelength can be selected with the external wavelength-selective reflector. To produce an optical short pulse, it can be operated at a gain-switching mode. This optical short pulse is blocked or passed by another RSOA #2 depending on bit 0 or bit 1 of the data signal based on external modulation technique. Then, the modulated optical short pulses pass through the N-chip Fourier encoder and are finally encoded and added into the ring. On the other hand, the OCDMA signals carrying on the dropped wavelengths are decoded by the N-chip Fourier decoder.
The generation process of the optical short train from the two RSOAs is described in detail here. An RSOA has inherently broadband spectrum and thereby a same RSOA can be used as different WDM optical sources. To produce a specific wavelength from the RSOA, it needs to be wavelength-locked by an external wavelength selective reflector with the same wavelength as the node. The wavelength-selective reflector is placed in front of the anti-reflective (AR) coated facet of the RSOA, here FBG. Then the RSOA #1 is self-injection locked to the reflected light with the selected wavelength. To generate a short pulse train, the RSOA #1 is gain-switched by a clock signal with the same repetition rate of the modulation signal. The pulse width is determined by the gain switching dynamics in the gain medium, i.e. the RSOA #1 in this case. It is possible to tune the pulse width by adding some dispersive medium like a chirp FBG or a dispersion compensation fiber after the short pulse generator. The final short pulse train whose wavelength is matched to the specific wavelength of the node is shown in the upper part of Fig. 2 . For the assignment of different wavelength, the Bragg wavelength of the FBG needs to be changed. Also, the wavelength shift caused by thermal deviation can be reduced by using athermal FBG with < 0.7 pm/Κ [13].
The Fourier code is a kind of complex-valued Hadamard code. This code can be obtained from the orthogonal property between the rows of the Fourier matrix F N which is well-known as the twiddle factor in the Cooley-Tukey fast Fourier transform (FFT) algorithm. The elements of the Fourier matrix F N are exp(2πj(m-1)(n-1)/N), where j is the imaginary unit, m and n are the positive integer indices for rows and columns from 1 to N. Equation (1) shows the Fourier matrix of N=4 as an example. Each row represents a 4-chip Fourier code as a QPSK code (C i). Each element of the matrix shown in Eq. (2) reflects the phase shifts between neighboring elements of the corresponding row of Eq. (1). Also, as we see in Eq. (2), the orthogonal code corresponding to the second row can be obtained by giving a phase shift of π/2 to the first orthogonal code and the other codes can be obtained in a same way without considering individual elements.
The Fourier encoder/decoder is made of four FBGs with the partial reflectivity separated from each other by the same distance and Bragg wavelength in a cascaded form as shown in Fig. 3 . The phase difference(ΔØ) between neighboring gratings is adjusted by the refractive index change or the Bragg wavelength as shown in Eq. (3), where L is the FBG spacing (5 mm in our experiment), n is the refractive index of the fiber core between two neighboring gratings, and λB is the Bragg wavelength of the grating. For each 4-chip Fourier code, one optical short pulse on single wavelength becomes encoded four optical short pulses reflected from four gratings in time domain and the encoded four optical short pulses have same phase difference between neighboring optical short pulses. In addition, to avoid the interference between the encoded neighboring optical short pulses, the pulse width should be less than the round trip optical delay (2ΔT=48.5-ps) between the chips as shown in Fig. 3. This type of encoder/decoder based on binary Hadamard code was reported to be stable on thermal variation by using a feedback controller as a thermal stabilizer [14].
The amount of phase shift is controlled by the spacing between FBGs or changing the operating wavelength for the given spacing. This wavelength-dependent phase shift induces a code mismatch and naturally acts as a WDM filter to the wavelength of the unwanted signal. In conclusion, the QPSK code using Fourier code provides lower cross-correlation property and easier code adaptation than those using binary Hadamard code like BPSK code [14]. Besides, this easy code adaption can be used for dynamic channel add/drop function because each code element is simultaneously controlled in the optical ring network.
3. Experiment
To evaluate the performance of the proposed ring, an experimental setup with two nodes was organized as shown in Fig. 4 . In general, multiple nodes should be considered. However, due to the limitation in the available encoders and decoders, we demonstrated with two wavelengths and four Fourier codes. In fact, there could be two types of ring. If we consider only drop case, each node can drop multiple wavelengths and multiple codes, multiple wavelengths and single code, single wavelength and multiple codes, or single wavelength and single code. For single wavelength and single code, we can construct eight nodes with two wavelengths and four codes. For higher capacity, we can increase the number of available wavelengths along the ring. Or, initially, we can design enough codes. Then we assign the codes to the SONET or Ethernet signals as many as we need leaving other codes for expansion. For tunable add and drop case, at each node, the specific wavelength is dropped with multiple codes and at the receiver side, we can select a particular code by tuning the decoder. For the case of signal add, if each node has only one code, then we can reroute a signal path from node i→node j to node i→node k by changing the encoder at the transmitter of node i. To show the proof of concept without the loss of generality, the wavelength spacing is set to 0.8 nm according to the ITU-T Recommendation for WDM channel allocation. Also, Fourier encoder and decoder acts as an optical filter for other wavelengths including ASE noise except for its own wavelength. Furthermore, the multiple interference noise from the other nodes due to cross-correlation can be further reduced by using Fourier code [9]. To show the feasibility of the proposed scheme, wavelength channel cross-talk, code interference, and add/drop function are investigated. Node 1 included Encoder#1 with the code word C 1 for λ1, i.e., (C 1, λ1). In addition, instead of tuning Encoder#1, Encoder#2 with the code word C 2 for λ1, i.e., (C 2, λ1) was used for investigating add/drop function and measuring code interference. On the other hand, Node 2 had Encoder#3 with the code word C 1 for λ2, i.e., (C 1, λ2) and Decoder#1 with the variable code C i for λ1, i.e., (C i, λ1: index i means a variable code, when i = 1, it shows auto-correlation with (C 1, λ1), cross-correlation with (C 2, λ1), WDM channel cross-talk with (C 1, λ2)). By changing C i from C 1 to C 4, add/drop function can be simulated. From Node 1, the generated pulses were injected into RSOA#2 through an optical circulator and RSOA#2 was wavelength-locked to that of the injected light as a λ1. Then, the modulation signal of 1.25 Gb/s was applied to RSOA#2 through a Bias-T. Depending on bit 1 or bit 0 of the modulation signal, the injected pulses remained or disappeared. Finally, the output of RSOA#2 showed a gapped short pulse train corresponding to the data pattern of the signal. This gapped signal passed through the circulator, split into two to be encoded by C 1 and C 2, and was optically amplified with an erbium doped fiber amplifier (EDFA) before being launched to the ring. The signal propagating through the single mode fiber (SMF) of 10.64 km along the ring is transferred to another EDFA at Node 2. This amplified signal is dropped and decoded by Decoder#1, simultaneously. The output was detected by using a photo-detector and amplified by a 4-GHz limited electrical amplifier [10]. For WDM channel cross-talk, optical short pulses with the wavelength of λ2 were encoded by Encoder#3, not modulated for simple demonstration and decoded by Decoder#1. The encoder/decoder fabricated with Fourier code has a code adaptation from C 1 to C 4 by the wavelength difference of about 0.15 nm on the operating wavelength. If we consider the typical WDM channel spacing of 0.8 nm, this change to the code characteristic can be neglected.
Fig. 4 Experimental setup. RSOA: reflective semiconductor optical amplifier, AMP: electrical amplifier, EDFA: erbium-doped fiber amplifier, SMF: single mode fiber, ISO: isolator, PPG: pulse pattern generator, PD: photo detector, BERT: bit error rate tester, DCA: digital communication analyzer
4. Results and discussion
Figure 5 shows the optical short pulses with the repetition rate of 1.25 GHz from RSOA#1 and RSOA#3 (biased at 18 mA and modulated by peak-to-peak voltage of 2 V), an optical spectra with optical power of −20 dBm and a side mode suppression ratio (SMSR) of about 30 dB. The Bragg wavelengths of the FBGs used in the short pulse generator of Node 1 and Node 2 were 1552.07 nm and 1552.82 nm, respectively. The reflectivity and pass band of the FBGs used at both nodes for self-injection locking were about 45% and 0.37 nm, respectively. Each optical short pulse has 32-ps pulse-width which satisfies the 48.5-ps chip interval of each encoder/decoder. This pulse width is good enough to generate a 4-chip code word at 1.25 Gb/s.
Fig. 5 Optical short pulse trains and their optical spectra: the signals at (a) pulse width: 32 ps, λ1: 1552.07 nm, (b) pulse width: 32 ps, λ2: 1552.82 nm
The modulated and encoded signals are shown in Fig. 6 . The bias current to RSOA#2 was 43 mA and the data pattern was ‘1011’ at 1.25 Gb/s. As we expected, the modulated signal shows gapped pulses with a missing pulse at the time of bit 0 as shown in Fig. 6 (a). Figure 6 (b) and (c) shows the modulated signals encoded by two different code words, C 1 and C 2, respectively. The ASE noise from RSOAs was effectively suppressed after passing through the encoders due to the filtering effect of the encoder/decoder. Figure 7 shows the wavelength-dependency of the encoder fabricated. It was displayed over the wavelength range of 1551.3 – 1552.8 nm. If the operating center wavelength is shifted by about 0.05 nm, then the code is changed from C 1 to C 2 (i.e., C 1 code to 1552.07 nm and C 2 to 1552.12 nm). This means that the wavelength channel spacing should be larger than 0.2 (0.05 × 4) nm to discriminate between neighboring orthogonal codes in WDM applications.
Fig. 6 (a) the pulse train modulated by the data pattern ‘1011’, (b) the modulation signal encoded by C 1, and (c) the modulation signal encoded by C 2
To investigate the contrast ratio of the OCDMA encoder/decoders, their auto-correlation peak (ACP) and cross-correlation peak (CCP) were measured and theoretically calculated by adjusting two optical switches at Node 1 and fixing C 1 at Node 2 as shown in Fig. 8 . In Fig. 8, left side represents experimentally measured results and right side represents theoretical results, respectively. The measured results were consistent with the theoretical results. The ratio of ACP over CCP appeared to be more than 5, respectively. These results prove that the Fourier encoder/decoder has a good contrast ratio [15]. Also, we did not use an additional WDM filter to separate the wavelength λ2 from Node 2 because other wavelengths except for λ1 were removed by the wavelength-dependent characteristic of the encoder/decoder.
Fig. 8 Measured (left) and theoretical (right) correlation waveforms after being decoded, (a) auto-correlation waveform for (C 1, λ1) × (C 1, λ1), (b) cross-correlation waveform for (C 2, λ1) × (C 1, λ1)
To simulate add/drop function, the outputs of Encoder#1 and #2 were combined at Node 1 and detected with the Decoder#1 by mechanically tuning C 1 to C 4 and their results are shown in Fig. 9 with theoretical results. In theoretical results of Fig. 9, their auto-correlation waveforms were not symmetrical for the center pulses. The difference between the experimental and theoretical results came from the phase error and the difference of timing. The theoretical results were obtained under the assumption of complete synchronization between two channels and simple Gaussian shape as the source pulse shape without red chirp caused by gain-switching mode of RSOAs. Under the existence of two encoded signals, the peaks were observed only for the cases of C 1 to C 1 and C 2 to C 2 and the ratios of ACP over CCP were also more than 5, respectively. These results show that the Fourier encoder/decoder can be used as a dynamic add/drop filter.
Fig. 9 Measured (left) and theoretical (right) correlation waveforms after code switching to Decoder#1, (a) auto-correlation waveform for (C 1+C 2, λ1)×(C 1, λ1), (b) auto-correlation waveform for (C 1+C 2, λ1)×(C 2, λ1), (c) cross-correlation waveform for (C 1+C 2, λ1)×(C 3, λ1), (d) cross-correlation waveform for (C 1+C 2, λ1)×(C 4, λ1)
Finally, we measured the code interference-induced power penalty and WDM channel cross-talk induced power penalty through their bit error rate (BER) curves and plotted them in Fig. 10 . From the BER curves, without code interference, each auto-correlation considering wavelength channel cross-talk shows almost the same BER curves. This means that the channel spacing of 0.8 nm does not make any channel cross-talk. By contrast, about 0.4-dB power penalty resulted with code interference. From the scalability point of view for higher number of users, the code-length of the Fourier code is the same with the maximum number of OCDM channels. The more the number of codes, the less the phase margin in the Fourier code. The code interferences caused by multiple Fourier codes can be reduced just by limiting the receiver bandwidth [10], instead of nonlinear signal processing technique. To measure the stability due to the drift in the operating wavelength, we gave a wavelength shift of 0.018 nm to λ1 at Node 1 and measured its BER. The corresponding power penalty appeared to be about 0.8 dB, and the wavelength drift within 0.018 nm was negligible. If an athermal FBG whose thermal dependency is less than 0.7 pm/Κ [14] is used as a reflector, this power penalty could be more improved.
Fig. 10 Measured BER curves, : BER for (C 1, λ1) × (C 1, λ1), : BER for (C 1, λ1) × (C 1, λ1) on wavelength shift of 0.018nm, : BER for (C 1, λ1 + λ2) × (C 1, λ1), : BER for (C 1 + C 2, λ1) × (C 1, λ1), : BER for (C 1 + C 2, λ1) × (C 2, λ1)
5. Conclusion
A hybrid WDM/OCDMA ring with a dynamic add/drop function was proposed and demonstrated. For the dynamic add/drop function, Fourier code with the same phase difference between the orthogonal codes was employed. For an optical source well adapted to WDM channels, RSOAs were used with FBGs and were gain-switched to generate the optical short pulse train for driving the encoder/decoder. The results of the experiment showed that the proposed ring has a power penalty of 0.4 dB between different orthogonal codes, and WDM channel cross-talk is negligible. Also, through the proposed ring, error free transmission over a fiber length of 10.64 km at a BER of <10−9 was achieved.
Acknowledgments
This work was supported by the National Research Foundation of Korea (NRF) Grant (NRF-2008-F01-2008-000-10012-0) and by the (Photonics 2020) research project through a grant provided by the Gwangju Institute of Science &Technology in 2011.
References and links
1. P. R. Prucnal, M. A. Santoro, and T. R. Fan, “Spread spectrum fiber-optic local area network using optical processing,” J. Lightwave Technol. 4(5), 547–554 (1986). [CrossRef]
2. P. J. Urban, B. Huiszoon, R. Roy, M. M. de Laat, F. M. Huijskens, E. J. Klein, G. D. Khoe, A. M. J. Koonen, and H. de Waardt, “High-Bit-Rate Dynamically Reconfigurable WDM–TDM Access Network,” IEEE J. Opt. Commun. Netw. 1(2), A143–A159 (2009). [CrossRef]
3. J.-F. Huang, Y.-T. Changa, and C.-C. Hsua, “Hybrid WDM and optical CDMA implemented over waveguide-grating-based fiber-to-the-home networks,” Opt. Fiber Technol. 13(3), 215–225 (2007). [CrossRef]
4. X. Wang, N. Wada, T. Miyazaki, G. Cincotti, and K. Kitayama, “Hybrid WDM/OCDMA for next generation access network,” Proc. SPIE 6783, 678328, 678328-14 (2007). [CrossRef]
5. X. Chen, G. Xia, D. Huang, and X. Yuan, “Experimental demonstration of 40 Gbit/s hybrid optical code-division multiplexing/wavelength-division multiplexing system,” Opt. Eng. 46(11), 115006 (2007). [CrossRef]
6. C. Zhang and K. Qiu, “Design and analysis of coherent OCDM en/decoder based on photonic crystal,” Opt. Lasers Eng. 46(8), 582–589 (2008). [CrossRef]
7. P. C. Teh, M. Ibsen, J. H. Lee, P. Petropoulos, and D. J. Richardson, “Demonstration of a four-channel WDM/OCDMA system using 255-chip 320-Gchip/s quarternary phase coding gratings,” IEEE Photon. Technol. Lett. 14(2), 227–229 (2002). [CrossRef]
8. S. Boztas, R. Hammons, and P. V. Kumar, “4-phase sequences with near-optimum correlation properties,” IEEE Trans. Inf. Theory 38(3), 1101–1113 (1992). [CrossRef]
9. M. Hanawa, “Fourier code: A novel orthogonal code for OCDM systems,” in Opto-Electronics and Communications Conference and Australian Conference on Optical Fibre Technology (OECC/ACOFT’ 2008), Sydney, 1–2 (2008).
10. M. Hanawa, “Multiple access interference reduction by limiting receiver bandwidth on Fourier code based-OCDM system,” in Opto-Electronics and Communications Conference (OECC’ 2009), Hong Kong, 1–2 (2009).
11. J. Wu and C.-L. Lin, “Fiber-Optic Code Division Add-Drop Multiplexers,” J. Lightwave Technol. 18(6), 819–824 (2000). [CrossRef]
12. C. -S. Bre's, I. Glesk, R. J. Runser, and P. R. Prucnal, “All-Optical OCDMA Code-Drop Unit for Transparent Ring Networks,” IEEE Photon. Technol. Lett. 17, 1088–1090 (2005). [CrossRef]
13. Advanced Optics Solutions Gmb, (http://www.aos-fiber.com/eng/FBG/Athermalen.html).
14. K. Hosoya, M. Hanawa, and K. Nakamura, “Programmable FBG-based variable optical correlator for optical code division multiplexing,” in Asia-Pacific Conference Communications (APCC’2009), Shanghai, 560–563 (2009).
15. N. Wada, “Optical Code Processing System, Device, and its Application,” JNW 5(2), 242–250 (2010). [CrossRef]
