Open Access
Add to BibSonomyAbstract
An optical thresholding technique based on super-continuum generation in dispersion flattened fiber is proposed and experimentally demonstrated to enable data-rate detection in optical code division multiple access networks. The proposed scheme exhibits an excellent discrimination between a desired signal and interference signals with features of pulse reshaping, low insertion loss, polarization independency as well as reasonable operation power.
©2005 Optical Society of America
1. Introduction
Optical code division multiple access (OCDMA) is one promising candidate for the next-generation broadband multiple access technique other than time division multiple access (TDMA) and wavelength division multiple access (WDMA) [1–4]. It has unique advantages of full asynchronous access, low latency owing to all-optical encoding/decoding, soft capacity on demand, protocol transparency, simplified network control, increased flexibility of QoS control as well as robust information security. Recently, coherent OCDMA using ultra-short optical pulse is receiving increasing interest due to the overall superior performance compared with the incoherent schemes and the progress of reliable and compact encoder/decoder devices, such as spatial light phase modulator (SLPM), planar lightwave circuit (PLC) and superstructured fiber Bragg grating (SSFBG) [5–12].
However, the coherent OCDMA system could suffer from severe multiple access interference (MAI) and beat noise, which limit the maximum number of active users that can be supported in a network [5]. One effective method to reduce the beat noise as well as the MAI noise is to lower the interference level by adopting ultra-long optical code (OC) [5]. For this purpose, recently, 511-chip long OC has been successfully demonstrated using superstructured fiber Bragg grating (SSFBG) at 640G-chip/s chip-rate [6–7]. On the other hand, in the coherent OCDMA with ultra-short optical pulse, the properly decoded signal is rather narrow (in a chip time) compared with the bit duration. In a practical system that employs “data-rate” instead of “chip-rate” detection, the MAI noise still remains to be a serious problem. Therefore, applying optical thresholding technique is crucial to enable data-rate detection for achieving a practical OCDMA system [5, 8].
Several optical thresholding techniques have been applied by using periodically-poled lithium niobate (PPLN) [8] and nonlinear effect in dispersion shifted fiber [9–10], high nonlinear fiber (HNLF) [11] and holey fiber [12]. So far, using the second-harmonic-generation (SHG) in PPLN has achieved the lowest operation power among them. However, the PPLN based device is polarization-dependant that will result in additional polarization mode partition noise in the system. Particularly, this issue could become very severe in an asynchronous OCDMA system since the beat noise is very sensitive to the polarization states of signal [5]. On the contrary, fiber based devices could have less polarization dependency. Nonlinear optical loop mirror (NOLM) was reported to be able to suppress the pedestal of the decoded pulse with low operation power [10], however, it might be not suitable for optical thresholding since the power transfer function of NOLM does not have a steep thrteshold characteristic. While, using self-phase-modulation (SPM) induced signal spectral broadening followed by long-pass-filter [9,11–12] could have better thresholding performance, however, still operate at relatively high power so far. In this paper, we will report a high performance optical thresholding technique by using super-continuum (SC) generation in normal dispersion-flattened-fiber (DFF) for high chip-rate coherent OCDMA.
2. Operation principle and experiment
SC generation by launching picosecond optical pulse into normal DFF has been demonstrated and applied as broadband light source for WDM systems [13–15]. The operation power could be rather low with the DFF. The proposed SC-based optical thresholder is shown in Fig. 1(a). It is composed of an EDFA, a 2 km-long DFF, and a 5nm bandpass filter (BPF). Figure 1(b) shows the dispersion characteristics of the DFF. The zero dispersion wavelengths of the fiber are 1523/1575 nm. The operation principle is that: the EDFA boosts the decoded optical signal to a proper level, the correctly decoded pulse, which have a well defined shape with ~2 ps pulsewidth and high peak power, will be able to generate SC in the DFF, while the incorrectly decoded signals (MAI noise) will spread over a large time span with very low peak power that is unable to generate SC. The BPF only allows the SC signal passing through and rejects the original signal. Therefore, after BPF, the correctly decoded signal can be recovered without MAI noise.
Fig. 1. (a) Configuration and operation principle of the SC-based optical thresholder (b) The dispersion characteristics of the DFF
Figure 2 shows the experimental setup for demonstrating the SC-based optical thresholding technique. In the experiment, the encoder/decoders were 511-chip, 640-Gchip/s SSFBGs. Each SSFBG en/decoder consists of 511 short (0.16-mm-long) FBG segments, between which, either 0 or π phase shifts are inserted according to some specific 511-chip binary-phase-shift-key (BPSK) Gold code [7]. An optical pulse train with ~1.8 ps pulsewidth was generated by the mode-locked laser diode (MLLD) and modulated by 223-1 pseudo-random bit stream (PRBS) data at 1.25 Gbit/s. The data stream was then split into two arms: Encoder 1 is the encoder of the desired user, while the MAI generator emulates the interferences from different undesired users. A fixed fiber delay line (FDL) was used in the signal’s path to de-coherence signal and interferences. Tunable optical delay line (TODL) was inserted in this path as well to investigate the impact of different phases of signal-interference overlapping. The MAI generator consists of 9 SSFBG encoders that can generate variable level interference. The time delay between two arms is adjusted to be longer than the coherent length of the optical signal. At the receiver, the decoded signal goes through the SC-based optical thresholder and detected by a photodetector (PD) with 30 GHz bandwidth. A lowpass filter (LPF) is placed to evaluate the performance with different receiver bandwidth.
The measured spectra of the original pulse, generated SC for different input powers and the signal after BPF are shown in Fig. 3(a) by the dashed grey line, solid grey lines and solid black line, respectively. The central wavelength of the BPF was set to be around 5 nm apart from that of the original signal. The measured power transfer function of this optical thresholder (Pout vs. Pin) is plotted in Fig. 3(b) using the correctly decoded signal without any interference. The operation power of the thresholder is about 1 dBm, which means the peak power of the signal user is about 6.3 W (~12.6 pJ/bit). This is higher than PPLN, but is one of the lowest among fiber-based optical thresholding techniques. The insertion loss (10log(Pin/Pout)) is about 13 dB at the operation point, which is also relatively low among the fiber based techniques. The measured transfer function exhibits a steep threshold characteristic compared to the sinusoidal transfer function of NOLM [10].
Fig. 3. (a) Measured spectra of the original pulse, generated SC with different input power and signal after BPF (b) Power transfer function of the SC-based optical thrtesholder
Figure 4(a) shows the waveforms of the decoded pulses with and w/o the SC-based optical thresholder measured by the second harmonic generation (SHG) autocorrelator. The red lines are with both correctly decoded signal and MAI noise from 9 interference users, while the black lines are with signal only. The signal and all the interferences were adjusted to have the same power in this case. The pulse widths of the original signals w/o and with MAI noise are measured to be 2.2 and 3.0 ps, respectively. Figures 4 (b) and (c) show the eye diagrams with and w/o SC-based optical thresholder, respectively. In this case, only one interference user has been used. The interference level to the desired signal was adjusted to around -24 dB and no LPF has been applied in the measurement. The significantly reduced noise level and clear eye opening verify the improvement of the optical thresholder. The regenerated optical pulse from the optical thresholder has a well defined shape with narrower pulse width of about 1.5 ps. The time-bandwidth product is ≅0.94, which means the regenerated pulse is closing to be transform-limited. This unique pulse reshaping function enables further all optical signal processing. The power contrast ratio [9] is measured around 13 dB. This is not so high due to the power leakage of the original signal from the BPF edge. Much improvement can be expected by using a sharp edge BPF with a bit longer central wavelength. Table I compares the overall performances of different optical thresholding techniques. This SC-based optical thresholding is characterized by the polarization-independency, rather low insertion loss, steep transfer function as well as the pulse reshaping capability.
Fig. 4. (a) SHG traces of the original pulses and sigals after optical thresholder (b) Eye diagram before the SC-based optical thresholder (c) Eye diagram after the SC-based optical thresholder
Table I. Performance comparison of different optical thresholding techniques
To investigate the impact of receiver’s bandwidth and demonstrate the improvement of using SC-based optical thresholder for data-rate detection, we measured the BER performances with and w/o this SC-based optical thresholder for different receiver bandwidth as shown in Fig.5. In Fig. 5, curves with circles are with LPF whose cutoff frequency fc~2.64 GHz (represents data-rate detection); curves with triangle marks are w/o LPF, therefore the bandwidth is limited by the BER tester (fc~12.5 GHz). Compared with those w/o optical thresholder, error free (BER‹10-9) has been attained with data-rate detection; while in the latter case, the receiver sensitivity is improved about 4 dB for BER=10-9 by using this optical thresholding technique. These results show that the impact of MAI noise increases with the decrease of receiver’s bandwidth and verify that data-rate detection for the decoded signal with MAI noise could be enabled by using the proposed optical thresholder. However, though the optical thresholder could only eliminate the MAI noise outside the auto-correlation peak, the signal-interference (SI) beat noise that accompany with the auto-correlation peak still remains and dominants the system performance [5, 16]. With a non-idea optical thresholder, the SI beat noise could be even worse. That might be the reason why in Fig. 5, the BER curves with SC optical thresholder have a relatively slower slope compared to that without thresholder.
Fig. 5. BER performance with and w/o the SC-based optical thresholder for different receiver bandwidth fc
We have also applied this optical thresholding technique in multi-user asynchronous OCDMA experiment. Up to ten active users has been successfully demonstrated in the experiment verifying the effectiveness of the proposed thresholding technique [16]. In the multi-user experiment, signals from 10 active OCDMA users with 511-chip BPSK Golde codes (code 1~10) are mixed in a truly asynchronous way and decoded by 4 different decoders (for code 1, 2, 4 and 9). The operating powers of the optical thresholder for different number of active users are plotted in Fig. 6 together with the theoretical calculation. In the calculation, the power of the interferences was taken into account as well as that of the target signal, therefore the thresholder’s operation power (Pin) increases with the increase of K. The interferences were assumed to have the same power with an average additional insertion loss of about -1.1dB compared to the target signal. This additional insertion loss is because that the target signal’s spectrum and the frequency response of the SSFBG decoder is matched with each other, while the interferences’ spectra is not matched with the decoder, therefore, will suffer from higher insertion loss. The value was obtained by averaging the measured powers of all the interferences for the four decoders with different codes in the experiment. The marks in Fig. 6 are the measured results for four different codes. The average operation power is about 1.4 and 10.3 dBm for 1 and 10 active users, respectively. The experimental results agree with the calculation very well.
3. Conclusion
We have proposed and experimentally demonstrated an optical thresholding technique based on SC generation in normal DFF for OCDMA application. This SC-based optical thresholding is characterized by the polarization-independency, rather low insertion loss, steep transfer function as well as the pulse reshaping capability. In the gigabit OCDMA experiment with chip-rate as high as 640 Gchip/s, the proposed optical thresholding technique enabled data-rate detection for the decoded signal with MAI noise. Further performance improvement can be expected by optimizing the fiber design, and using a BPF with sharper edge.
Acknowledgements
The authors would like to thank the anonymous reviewers for their careful reading and valuable comments. The authors would also thank Y. Tomiyama, H. Sumimoto, and T. Makino of NICT for their technical support in the experiment.
References and links
1. I. Sankawa, “Fault location technique for in-service branched optical fiber networks,” IEEE Photon. Technol. Lett. 2, 766–1768 (1990). [CrossRef]
2. Y. Koyamada, T. K. Horiguchi, and S. Furukawa, “Recent progress in OTDR technologies for maintaining optical fiber networks,” in Tech. Dig., IOOC’95 (Hong Kong, 1995), FA1-4.
3. C. K. Chen, F. Tong, L. K. Chen, J. Song, and D. Lam, “A practical passive surveillance scheme for optical amplified passive branched optical networks,” IEEE Photonics Technol. Lett. 9, 526–528 (1997). [CrossRef]
4. C. K. Chen, F. Tong, L. K. Chen, K. P. Ho, and D. Lam, “Fiber-fault identification for branched access networks using a wavelength-sweeping monitoring source,” IEEE Photonics Technol. Lett. 5, 614–616 (1999).
5. A. D. Kersey and W. W. Morey, “Multi-element Bragg-grating based fiber-laser strain sensor,” Electron. Lett. 29, 964–966 (1993). [CrossRef]
6. C. H. Yeh, C. C. Lee, and S. Chi, “A tunable s-band erbium-doped fiber ring laser,” IEEE Photonics Technol. Lett. 15, 1532–1534 (2003).
7. C. H. Yeh, C. C. Lee, C. Y. Chan, and S. Chi, “S-band gain-clamped erbium-doped fiber amplifier by using optical feedback method,” IEEE Photonics Technol. Lett. 16, 90–92 (2004). [CrossRef]
