1. Introduction

Multi-access interference (MAI) is one of the key factors limiting the performance of optical code division multiple access (CDMA) systems [1]. Therefore, discrimination of autocorrelation peaks is an essential element in optical CDMA receivers which can improve the detection of optical CDMA signals by rejecting MAI [2]. Autocorrection peak discrimination is also an important function for code dropping in passive optical bus or ring networks. To suppress MAI, an optical thresholder [3–7] is commonly used, especially when the MAI is small compared to the autocorrelation peak. Typical optical thresholders based on nonlinear phase shifts [3] require the input signal to have a sufficiently large amplitude and narrow pulse width to produce the phase shift needed to switch the autocorrelation peaks.

In this paper, we propose and experimentally demonstrate a novel optical autocorrelation peak discriminator based on four-wave mixing (FWM) in a 37-cm highly-nonlinear bismuth oxide fiber (Bi-NLF). It is experimentally shown that with the use of the FWM-based autocorrelation peak discriminator, signal detection of a 2-D incoherent optical CDMA system is substantially improved. The Bi-NLF has a very high nonlinear coefficient of 1100 W⁻¹km⁻¹ that makes our approach very compact and have a low latency. Unlike previous optical thresholding techniques, FWM does not require a fast change in signal intensity to induce the nonlinear effect, therefore it is not necessary for the input to have a narrow pulse-width. This discriminator does not impose any requirement on the power ratio of the autocorrelation peak and cross-correlation peak, so the scheme works well even when the MAI has a higher power than the autocorrelation peak. In this paper, the FWM-based autocorrelation peak discriminator is used to improve optical CDMA detection by rejecting MAI. Eye diagrams are measured, and it is shown that the autocorrelation peak is successfully extracted from a received signal with MAI, and error-free performance is obtained. Optical CDMA signal detection improvements under different power ratios are studied, and bit-error rate (BER) measurements are performed. An improvement in receiver sensitivity of more than 2 dB is obtained. In addition, error floors are removed with the use of the FWM-based autocorrelation peak discriminator.

2. Principle and experimental setup

The proposed autocorrelation peak discriminator exploits FWM in a 37-cm Bi-NLF. The operation is based on the fact that FWM occurs only when the input wavelengths are aligned in time, which is a condition that is satisfied by the autocorrelation peaks. In a 2-D optical CDMA system which uses wavelength-hopping time-spreading codes [8], an optical correlator typically decodes the signal by aligning all the wavelengths of the desired code in time, while all the wavelengths of the interfering users are spread over time. The operation of the optical CDMA decoder results in an autocorrelation peak for the desired code, and MAI for all other codes, as illustrated in Fig. 1(a) . Since the wavelengths of the MAI are not aligned in time, it cannot produce FWM, even if the total power level is high. Therefore, FWM occurs only for the autocorrelation peaks, which necessarily have all the wavelengths of the code aligned in time. By selecting the FWM output, only the autocorrelation peaks are obtained, as illustrated in Fig. 1(b).

 

Fig. 1 Schematic illustration. (a) Generation of autocorrelation and cross-correlation peaks during optical CDMA signal decoding. (b) Optical spectrum of the FWM based autocorrelation peak discriminator.

Download Full Size | PDF

Figure 2 shows the experimental setup of the autocorrelation peak discriminator. To demonstrate MAI suppression at a 2-D incoherent optical CDMA receiver, two encoded wavelength-hopping time-spreading optical CDMA signals are simultaneously present in the received signal. The coding scheme used is a carrier-hopping prime code [9], where each user is assigned a unique code consisting of all the wavelengths with each one located in a unique time slot within a bit. The received signal is launched to an optical CDMA decoder matching the code of “user 1.” At the decoder output, an autocorrelation peak results from user 1, while user 2 appears as MAI. To improve signal detection in the optical CDMA system, the autocorrelation peak has to be discriminated from the MAI in the decoded received signal. The decoded received signal consists of the data from user 1 that forms the autocorrelation peaks, and the MAI from user 2. The signal is then amplified and launched into the FWM-based autocorrelation peak discriminator. FWM takes place in the Bi-NLF only when the autocorrelation peak presents. Since the wavelengths of the MAI are not aligned in time, no FWM can occur, even if the power is high.

 

Fig. 2 Experimental setup of the autocorrelation peak discriminator. MAI: Multi-access interference; Bi-NLF: Highly-nonlinear bismuth oxide fiber; BPF: Optical bandpass filter.

Download Full Size | PDF

In our experiment, the optical CDMA signals are generated through pulse carving of four cw DFB laser output by an electro-absorption modulator. The signals are then modulated with a 2³¹-1 pseudo-random binary sequence (PRBS). To encode the modulated signal, two different fiber Bragg grating-based encoders (encoder 1 and encoder 2) are used. Thus, two optical CDMA signals are generated corresponding to user 1 and user 2. The carrier hopping prime code used in the experiment consists of four wavelengths and seventeen time chips in each bit. The two codes we use are (1, 4, 7, 10) for user 1 and (1, 8, 15, 5) for user 2. Hence, the 1st, 4th, 7th, and 10th time chips of user 1 and the 1st, 8th, 15th, and 5th time chips of user 2 are filled by four pulses with different wavelengths. After encoding, the four wavelengths in the optical CDMA signals are spread in time according to the code design in the encoder. To facilitate the study of the autocorrelation peak discrimination scheme, the power of user 1 is fixed while that of the user 2 is tunable. The two signals are combined at the coupler before transmission. At the receiver, a signal consisting of user 1 and user 2 is launched into decoder 1 for the detection of user 1 data. An autocorrelation peak is formed from user 1, while a cross-correlation signal is generated by user 2. By amplifying the decoded output and launching it into a 37-cm Bi-NLF with nonlinear coefficient of 1100 W⁻¹km⁻¹, FWM occurs for the autocorrelation peaks only. Since no FWM occurs for MAI, the information is obtained only from the autocorrelation peaks by selecting the FWM output using an optical bandpass filter. In this way, discrimination of the autocorrelation peak is achieved that aids in the detection of the optical CDMA signal.

3. Results and discussion

The optical CDMA signals consist of four wavelengths with 0.8 nm channel spacing at 1550.12 nm, 1550.92 nm, 1551.72 nm, and 1552.52 nm. The power of user 1 is 0 dBm, and that of user 2 is tuned from −3 dBm to + 3dBm to study the performance of this scheme. By modulating the optical CDMA signals with 2³¹-1 PRBS signal, eye diagrams of the decoded signals are measured with user 2 to user 1 power ratios of 3 dB, 0 dB, and −3 dB, as shown in Fig. 3(a) , Fig. 3(c), and Fig. 3(e), respectively. Very strong cross-correlation peaks are observed in the 3-dB case. The decoded signals are amplified and launched into the 37-cm Bi-NLF to perform FWM-based autocorrelation peak discrimination. The FWM output is selected using an optical bandpass filter and only autocorrelation peaks are obtained. The corresponding eye diagrams are shown in Fig. 3(b), Fig. 3(d), and Fig. 3(f). It is clearly seen that all the cross-correlation peaks from user 2 have nearly vanished, enabling the detection of user 1 to be improved. A small residual MAI is observed in Fig. 3(b) and Fig. 3(d), resulting from imperfect filtering of the FWM output.

 

Fig. 3 Eye diagrams of decoded signal with different user 2 to user 1 power ratios (a) 3 dB (b) 3 dB with autocorrelation peak extraction (c) 0 dB (d) 0 dB with autocorrelation peak extraction (e) −3 dB (f) −3 dB with autocorrelation peak extraction.

Download Full Size | PDF

To study the improvement of optical CDMA signal detection using our autocorrelation peak discriminator, BER measurements are performed at different power ratios between the two users, as shown in Fig. 4 . The hollow data points correspond to BER measurements made right after decoding, while the solid data points are the measurements after autocorrelation extraction using FWM. For a relatively small user 2 to user 1 power ratio of −1 dB, a 2-dB receiver sensitivity improvement is obtained. For a large power ratio of 1 dB, the receiver sensitivity is improved and the error floor is removed. For a power ratio of 3 dB, the MAI is too strong, and we are not able to measure the BER of the decoded signal. However, after the discriminator, error-free performance is sucessfully obtained.

 

Fig. 4 BER measurement at different user 2 (cross-correlation) to user 1 (autocorrelation) power ratios. Hollow: Before autocorrelation peak extraction. Solid: After autocorrelation peak extraction. For a power ratio of 3 dB, the MAI are too strong and the BER of the decoded signal cannot be measured without the autocorrelation peak discriminator.

Download Full Size | PDF

It is worth noticing that the proposed autocorrelation peak discriminator works well even when the cross-correlation peaks are of equal or much higher amplitude than the autocorrelation peaks. This is a consequence of the fact that FWM occurs only when the wavelengths are aligned in time. To demonstrate this phenomenon, we have adjusted the powers of user 1 and user 2 such that some of the autocorrelation peaks and cross-correlation peaks have the same amplitudes. Figure 5(a) and Fig. 5(b) shows the eye diagram of the decoded signal before and after the autocorrelation peak discriminator. A 4-nm optical bandpass filter is used to block the original optical CDMA wavelengths, while the FWM output is obtained at the through-port of the optical bandpass filter. As shown in Fig. 5(b), the cross-correlation peak vanishes at the output. The small residual cross-correlation peaks resulted from imperfect filtering, and can be improved using a higher order optical bandpass filter with sharper band edges. By further increasing the amplitude of the cross-correlation peaks to become higher than the autocorrelation peaks, as shown in Fig. 5(c), it is seen that the proposed autocorrelation peak discriminator works well in extracting the autocorrelation peaks, as depicted in Fig. 5(d).

 

Fig. 5 Eye diagrams of decoded signal (a) some cross-correlation peaks have the same amplitudes as the autocorrelation peaks (b) cross correlation peaks nearly varnish after the discriminator (c) some cross-correlation peaks are of higher amplitudes than the autocorrelation peaks (d) cross correlation peaks nearly vanish after the discriminator.

Download Full Size | PDF

4. Conclusion

We propose and demonstrate an optical autocorrelation peak discriminator using four-wave mixing in a 37-cm highly nonlinear bismuth oxide fiber. Owing to the large Kerr nonlinearity of Bi-NLF, a very short length of fiber is sufficient to introduce significant FWM, providing a more compact solution than previous threshold detection approaches. The principle of the autocorrelation peak discriminator is based on the fact that FWM occurs only when the wavelengths are aligned in time, which uniquely occurs in the autocorrelation peaks. In contrast, no FWM occurs for the MAI, since its wavelengths are spread over time. This scheme has the additional advantage, compared to previous threshold-based optical CDMA signal detection schemes, that the autocorrelation peaks are readily extracted in the presence of large MAI. Eye diagrams and BER of the decoded optical CDMA signal are measured at different power ratios. The results show significant improvements in the receiver sensitivity, and the removal of error floors. We experimentally demonstrate that the scheme works well even when the cross-correlation peaks are of equal or higher amplitude than the autocorrelation peaks.