A novel scheme for suppressing the multiple-access interference (MAI) in coherent time-addressed optical CDMA systems is proposed. This is based on a differential detection using the dual-control NOLM. The basic principle for MAI suppression is described. For experimental demonstration, two encoded channels are constructed and decoded. These decoded signals are sent to the dual-control NOLM and a high autocorrelation peak with suppressed MAI at the output of NOLM is observed. Signal-to-interference ratio is improved by 7 dB.
©2004 Optical Society of America
Lately, optical code-division-multiple access (CDMA) systems are receiving great attention due to their high asynchronous connectivity and security . These features are very important for further system upgradability and possibility into optical access networks. Based on this, a time-addressed coding method, which spreads a pulse over data bit-period and receives it with the decoder matched to the encoder, is one of the popular schemes that have been frequently incorporated by many optical CDMA systems. In most cases, the most dominant source limiting the entire system performance is known as multiple-access interference (MAI), which originates from the cross-talks of the other users. In incoherent optical CDMA systems [1–3] that perform decoding only on power basis, the effects of MAI become more significant as the number of users increases. On the other hand, coherent optical CDMA systems [4–7] are more attractive because they perform the decoding on phase basis, so that a high central autocorrelation peak with low side lobes could be achieved. This is attributed to the coherent superposition between pulses during the decoding process.
As a method to reduce MAI, a time-gating scheme has been demonstrated  based on previous channel demultiplexing techniques in an optical time-division multiplexing (OTDM) system. However, the MAI existing in the same chip-time slot with the desired central autocorrelation peak cannot be eliminated by this time-gating scheme. In another method, there is a differential detection scheme employing the balanced photo-detector that rejects the common mode of two optical signals launched to the detector. In a 1-channel coherent matched decoding system , the balanced photo-detector has been used to suppress MAI. The encoding/decoding in this system are based on concatenated optical delay lines and 2x2 3-dB optical couplers, referred to as a ladder network. This ladder decoder has two output ports. When the decoder is exactly matched to the encoder (desired channel), the decoder provides the different waveforms in the central chip-time slot (i.e., high autocorrelation peak, intensity null). On the contrary, the signals from the other channels (i.e., MAI) appear with the same waveform at the output of decoder. With the balanced photo-detector, therefore, MAI can be rejected, while the desired autocorrelation peak passes through without significant loss. However, the whole operating speed in this scheme is limited by that of the used photo-detector, in fact, which is not fast enough to accommodate the rapidly increasing code speed (i.e., a few hundreds of Gchip/s ). Therefore, to overcome such speed limitation radically, an investigation for a new all-optical differential detection scheme with high processing speed is considered necessary.
In this study, a novel all-optical method for suppressing the MAI in coherent time-address optical CDMA systems is proposed and demonstrated. This MAI suppression was accomplished by an all-optical differential detection based on a dual-control nonlinear optical loop mirror (DC-NOLM). The basic principle of MAI suppression was described and its viability was experimentally demonstrated with two channels. With DC-NOLM, a signal-to-interference ratio was improved by 7 dB.
A schematic for coherent time-addressed optical CDMA systems with n-stage ladder networks is shown in Fig. 1. M channels consisting of a desired channel and (M-1) interfering channels are encoded with their unique sequence codes and transmitted simultaneously. At each receiver, these multiplexed signals are received through a decoder whose impulse response is matched to one of the sequence codes of M channels. An optical pulse is split with a pre-specified delay at each stage of the ladder encoder so that an address code with 2n pulses is constructed. When an optical pulse passes through a 2x2 coupler, it undergoes the phase shift of either 0 or π/2 radian, depending upon whether it goes straight or crosses the coupler. This phase shift therefore causes much more complex phase-based interferences during the decoding process .
Figures 2(a) and (b) show the timing view diagrams at two output ports (A and B) of the decoder. When the encoder/decoder has an identical network, one port (A) of the decoder provides an autocorrelation peak of p2; p = 2n while another port (B) has an intensity null in the same chip-time slot. The details of this coherent coding theory are described in Refs. . On the other hand, when the encoder/decoder networks are different from each other, the same output is provided in waveform and power, called MAI.
The two outputs of the decoder are sent to two control ports of DC-NOLM. This DC-NOLM transmits the power corresponding to the differential power between the control input signals. This operation of DC-NOLM has been used to implement an exclusive-OR (XOR) logic gate . A schematic illustration of DC-NOLM and the truth table of XOR logic are shown in Fig. 3. The XOR logic is characterized by providing a logic 1 output only if two input logics (LA and LB) are different. If both input logics are the same, the gate provides a logic 0. DC-NOLM is composed of an optical coupler, a fiber loop with an optical Kerr medium, and two wavelength-division-multiplexers (WDMs) for introducing the control signals into the loop. When a continuous wave (CW) probe lightwave enters DC-NOLM, it is divided into two equal counter propagating fields. As it travels through the loop, each lightwave experiences a cross-phase modulation (XPM)-induced nonlinear phase shift by the control signals, given as
where P(t) is the power of control input signal, Pπ is the π-phase shifted power, and Leff is the effective fiber length. After being coupled again through the coupler, they are interfered with a relative phase difference Δϕ(t) between the clockwise and counterclockwise lightwaves . Assuming that the power coupling ratio of the optical coupler is 0.5, an output transmittance from DC-NOLM is given by
In the absence of control input signals to DC-NOLM (LA=LB= logic 0), Δϕ(t) becomes zero so that two counter propagating probe lightwaves are destructively interfered and all input lightwaves are reflected from DC-NOLM. Similarly, even when both control signals are launched with equal powers (LA=LB= logic 1), Δϕ(t) becomes zero because the same amount of nonlinear phase shifts occurs to two probe lightwaves, as in the case of no control input. This corresponds to the case when the MAI with same waveform is received. No power is transmitted to the output of DC-NOLM. On the contrary, when a control signal is launched on only one of two ports (either LA or LB = logic 1), the phase balance is broken and some power of probe lightwave is transmitted. This corresponds to the case when the desired signal with imbalanced power is received. Therefore, it can be seen that DC-NOLM acts an all-optical differential detector that rejects the MAI in common mode.
3. Experiment and results
Experimental demonstration was carried out using the experimental set-up shown in Fig. 4. Optical input pulses were generated from a laser diode (LD1), gain-switched by RF sinusoidal signal at 1.25 GHz, 27 dBm, providing 10 ps (FWHM) pulse trains centered at 1552 nm wavelength. This pulse source was amplified by an erbium-doped fiber amplifier (EDFA1) and divided equally, then sent to the encoder1 (desired channel) and encoder2 (interfering channel) with delays of 130 ps and 200 ps, respectively. It should be noted that in practice, the encoder and the decoder are placed separately at a remote location. Therefore, it is a prerequisite to adjust the delay network of the decoder to be matched to that of the encoder within a coherence time of input pulse source (~a few hundreds of fs)  to ensure a coherent matching. Also, the decoder should be able to compensate for the ambient temperature change in optical fibers since temperature change may cause severe phase difference between the two networks. Fortunately, this problem could be remedied by incorporating active phase-locking circuits composed of electrical feedback loops and dc phase shifters at each delay line .
This study focused on the proof-of-principle regarding MAI reduction by DC-NOLM. Thus, the encoder1 was reused as its own decoder, reflecting the encoded pulses by placing an optical mirror at its output. Such reflected pulses undergo an identical path with the encoding process. Therefore, when those pulses are recombined, they must be coherently interfered so that two output ports (OC2, OC3) of the decoder will appear in the complementary states in the central chip-time slot. Their relative intensities are expected as (1 0 4 0 1) and (1 0 0 0 1), respectively.
On the other hand, an interfering channel was constructed with the encoder2, whose output pulses were amplified and coupled with the desired channel through an optical coupler (OC1). Passing the decoder, the interfering channel would be time-spread with the same output waveforms. Two signals at OC2 and OC3 were sent to the control ports of DC-NOLM, adjusting their polarization state and synchronizing time position using a variable optical delay line (ODL1). The CW probe lightwave was generated from a laser diode (LD2) centered at 1532 nm. The dispersion-shifted fiber (DSF) used in this experiment has a zero-dispersion wavelength of 1542 nm and length of 20 km. An optical bandpass filter (OBF) was used to select the converted probe lightwave. Final waveforms were observed by using a 20-GHz photo-detector and a sampling oscilloscope.
Figure 5 compares the spectra before (dotted line) and after (solid line) DC-NOLM. This shows the amplitudes of the two lines normalized to each other. Two peaks were observed at center wavelengths of 1532 nm (CW probe lightwave) and 1552 nm (control input signal). After DC-NOLM, the spectrum at 1532 nm was particularly broadened due to the XPM by the control signals at 1552 nm.
Measured temporal waveforms before and after DC-NOLM are shown in Figs. 6, 7, and 8. Figure 6 shows the case in which only the desired channel is present. As expected, a central autocorrelation peak (Fig. 6(a)) and an intensity null (Fig. 6(b)) were observed. After DC-NOLM, only the main lobe term remained while the other side lobes were removed. Fig. 7 shows the case in which an interfering channel is present without the desired channel. Since the delay network of decoder was not matched to that of the interfering channel, the outputs of decoder were time-spread with the same waveforms (Figs. 7(a) and (b)). After DC-NOLM, their powers were reduced remarkably (Fig. 7(c)). Fig. 8 shows the case in which both channels are simultaneously present. Before DC-NOLM (Fig. 8(a)), the peak intensity of the interfering channel was 1/4 for the autocorrelation peak of the desired channel. In this case, a signal-to-interference ratio (SIR) is defined as SIR = 10log10(Is/Ii) where Is and Ii are peak intensities of the desired channel and the interfering channel after decoding, respectively. This signal was found to have SIR of 6 dB. After DC-NOLM, SIR was improved up to 13 dB as other side lobes were reduced greatly. This corresponds to the SIR improvement of 7 dB. This result is a clear evidence proving the potential of the scheme presented in this study to remove MAI and to improve system performance in optical CDMA systems.
Note that in previous experiments, the power of interfering channel did not overlap or interfere with the central autocorrelation peak of the desired channel. In realistic systems, however, much more interfering channels transmitted asynchronously exist so that the individual interfering power falls and overlaps onto random timing position in a bit period. To simulate this effect experimentally, the time delay of the interfering channel over 200 ps (from -100 ps ~ +100 ps) was varied using an ODL2. The SIR was then measured as a function of the relative time delay between two channels (Fig. 9). The dotted line indicates the SIR (6 dB) before DC-NOLM. Along with the variation of the timing position, no significant discrepancy in SIR was observed.
Figure 10 shows SIR variation at the output of DC-NOLM as a function of autocorrelation peak power (Pc) of the desired channel. It shows an optimal control power condition that should be launched to the control ports of DC-NOLM. DC-NOLM had a maximum transmittance when Pc = 180 mW; this power was denoted as Pmax. However, the best SIR of 13 dB was observed at Pc = 120 mW (not Pmax). This was different from the expectation that the best SIR may appear when Pc = Pmax. Note that in this experiment, the cancellation for the interfering power (i.e., MAI) was not perfect. This imperfect cancellation was because the power of two control signals launched to the DSF could be slightly different. This could be probably due to uneven propagating loss between two lightwave paths from the decoder to the DSF. This unevenness might make Δϕ(t) to be nonzero. As a result, the amount of the interfering power transmitted to the output of DC-NOLM increased along with Pc.
Inspecting the waveforms at the output of DC-NOLM, it was observed that as Pc exceeded the optimal value (120 mW), the interfering power increased more rapidly than the signal power does. This is the reason why the best SIR was observed at Pc = 120 mW. Insets show corresponding waveforms when Pc’s are 120 mW and 220 mW, respectively. In the case of Pc = 220 mW, appreciable interfering power, which degraded SIR, could be observed. Therefore, in the system having imperfect MAI cancellation, an optimal input peak power (<Pmax) for the best SIR might exist. In addition, a trade-off considering the interfering power might be needed. Conversely, assuming that regardless of its input power, the perfect MAI cancellation can be achieved, the best SIR might appear when Pc ≈ Pmax. To improve SIR, the cancellation efficiency for MAI should be increased. This is another issue that requires further study.
The biggest advantage of the proposed method is its high operation speed due to the all-optical configuration based on NOLM. This was one of the distinguishing features from the methods based on the previous balanced photo-detector. This study offers a solution for the speed problem at the receiver.
Using a sequence code consisting of two pulses, the side lobes (except for central autocorrelation peak) of the desired signal after decoding were power-balanced to be cancelled. However, in most practical implementations using the longer sequence code, the side lobes at two ports of decoder are not exactly balanced. For a 2-stage ladder encoder/decoder in Ref. , their relative intensities were calculated as (1 0 1 16 1 0 1) and (1 4 1 0 1 4 1), respectively, and the differential output would be (0 4 0 16 0 4 0). To remove the remaining side lobes, a time-gating is required. If a pulse-type probe signal, synchronized to the central autocorrelation peak of the desired signal, would be used, then only the main lobe of the desired signals could be selected. This was another advantage the scheme presented in this study, along with the high-speed operation ability.
However, there were two troublesome impairments in this experiment. One was the instability of DC-NOLM that originated from the polarization birefringence and the long fiber length. This is a long standing problem in the research field with respect to NOLM. A number of works have been devoted to suppress such problems by employing a polarization maintaining fiber and a highly-nonlinear dispersion-shifted fiber (HNL-DSF) . Another impairment was the finite cancellation efficiency for MAI. This was due to the imperfect power-balance between two control input signals to DC-NOLM. Therefore, further study to improve the power uniformity between the propagating paths of two control input signals, including a planer-lightwave-circuits (PLCs) type encoder/decoder , is necessary.
A novel all-optical differential detection scheme was demonstrated using DC-NOLM for suppressing MAI in coherent time-addressed optical CDMA systems. The basic principle of MAI suppression was described and a corresponding experimental demonstration was successfully performed. Two channels were encoded with 1-stage ladder networks and decoded with one of such encoders to ensure a coherent matched filtering between decoded pulses. Signal waveforms before and after DC-NOLM were compared, finally a high central autocorrelation peak with suppressed MAI was observed. As a function of relative time delay between two channels, SIR variation at the output of DC-NOLM was measured. No appreciable SIR degradation was observed. In addition, the autocorrelation peak power of the desired channel was varied and the best SIR of 13 dB was observed at lower peak power (120 mW) than Pmax (180 mW).
This work was partially supported by KOSEF through grant No. R01-2001-000-00327-0 from the Basic Program and a Brain Korea 21 Project.
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, 547–554 (1986). [CrossRef]
2. J. A. Salehi, “Code division multiple-access techniques in optical fiber networks-Part I: Fundamental principles,” IEEE Trans. Commun. 37, 824–833 (1989). [CrossRef]
3. A. S. Holmes and R. R. A. Syms, “All-optical CDMA using “Quasi-prime” codes,” J. Lightwave Technol. 10, 279–286 (1992). [CrossRef]
4. Y. L. Chang and M. E. Marhic, “Fiber-optic ladder networks for inverse decoding coherent CDMA,” J. of Lightwave Technol. 10, 1952–1962 (1992). [CrossRef]
5. M. E. Marhic, “Coherent optical networks,” J. of Lightwave Technol. 11, 854–864 (1993). [CrossRef]
6. H. Sotobayashi, W. Chujo, and K.-I. Kitayama, “1.6-b/s/Hz 6.4-Tb/s QPSK-OCDM/WDM (4OCDM × 40WDM × 40Gb/s) transmission experiment using optical hard thresholding,” IEEE Photon. Technol. Lett. 14, 555–557 (2002). [CrossRef]
7. N. Wada and K-I. Kitayama, “A 10 Gb/s optical code division multiplexing using 8-chip optical bipolar code and coherent detection,” J. of Lightwave Technol. 17, 1758–1765 (1999). [CrossRef]
9. R. A. Griffin, D. D. Sampson, and D. A. Jackson, “Demonstration of data transmission using coherent correlation to reconstruct a coded pulse sequence,” IEEE Photon. Technol. Lett. 4, 513–515 (1992). [CrossRef]
10. S. Kim, T. Eom, B. H. Lee, and C. Park,“Optical temporal encoding/decoding of short pulses using cascaded long-period fiber gratings,” Opt. Express 11, 3034–3040 (2003), http://www.opticsexpress.org/abstract.cfm?URI=OPEX-11-23-3034. [CrossRef] [PubMed]
11. M. Jinno and T. Matsumoto, “Nonlinear sagnac interferometer switch and its applications,” IEEE J. of Quantum Electron. 28, 875–882 (1992). [CrossRef]
12. M. Jinno, “Effects of crosstalk and timing jitter on all-optical time-division demultiplexing using a nonlinear fiber sagnac interferometer switch,” IEEE J. Quantum Electron. 30, 2842–2853 (1994). [CrossRef]
13. Mohammed N. Islam, Ultrafast Fiber Switching Devices and Systems (Cambridge University Press, New York, USA, 1992).
14. S. Watanabe and S. Takeda, “All-optical noise suppression using two-stage highly-nonlinear fibre loop interferometers,” Electron. Lett. 36, 52–53 (2000). [CrossRef]