a745_1.pdf CMJJ1.pdf
Optimization of Receiver Filter Bandwidth for Externally Modulated Optical MSK Data Jinyu Mo1,2, Yang Jing Wen1, Yixin Wang1, Chao Lu3, and Wen-De Zhong2 1
Network Technology Department, Institute for Infocomm Research, A*STAR, 21 Heng Mui Keng Terrace, Singapore 119613 2 School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore 3 Department of Electronic and Information Engineering, Hong Kong Polytechnic University, Hong Kong Email:
[email protected]
Abstract: This paper examines the optimum optical and electrical filter bandwidths of the optical MSK receiver. Results show that the MSK signal with low OSNR exhibits narrower optimal optical filter bandwidth compared with RZ-DPSK and RZ-OOK. @2007 Optical Society of America OCIS codes: (060.4080) Modulation; (060.4510) Optical communications
1. Introduction Performance optimization of direct detection receivers in the presence of optical noise for high speed transmission systems has attracted great research attention [1-4]. Boivin et al [4] studied the optimum filter bandwidth for nonreturn-to-zero (NRZ) and RZ signal with arbitrary duty cycle. Winzer et al [2] evaluated the impact of filtering in a RZ differential phase shift keying (RZ-DPSK) system. Using optimal receiver design, the receiver sensitivity can be improved, which alleviates the requirement for other system parameters and/or extends transmission distance. In this paper, we optimize the receiver optical and electrical filter bandwidths for an externally modulated optical minimum shift keying (MSK) signal that we implemented recently [5]. The optimum optical and electrical filter bandwidths for the RZ-OOK (50% duty cycle) and RZ-DPSK (50% duty cycle) are also presented here, as a comparison for the MSK data. 2. System Modeling The schematic for receiver bandwidth optimization is shown in Fig.1. The MSK data was generated using the same implementation as described in [5]. The RZ-OOK and RZ-DPSK were generated using two Mach-Zehnder modulators (MZMs), with the pulse carving MZM biased at the quadrature point with Vπ driving swing. The optimum filter bandwidth study here was carried out for a 10 Gb/s system under two cases, the back-to-back (BTB) and after 8x80km transmission, via simulation using the commercial software, VPItransmissionmaker. Each span consists of 80km single mode fiber (SMF) and corresponding dispersion compensation fiber (DCF) to make full dispersion compensation. Two erbium doped fiber amplifiers (EDFAs) were used to compensate for the total power loss in each span, which were located before and after the DCF. Amplified spontaneous emission (ASE) noise was included to study the effects of different optical signal-to-noise EDFA1 EDFA2 EDFA ratios (OSNR). The OSNR could be changed by the attenuator 80km DCF (Att2). For all the formats evaluated, the optical signal with and SMF nd SW2 without transmission, passed through a 2 order Gaussian optical 3dB bandpass filter, with a bandwidth of Bo. For phase modulated Transmitter SW1 formats, an one-bit delay interferometer (DI) was used for data Att1 Att2 MSK, or RZ-OOK, demodulation, followed by a balanced receiver. For RZ-OOK, no or RZ-DPSK Bo DI and only single end receiver were used. An electrical 4th order DI Bessel lowpass filter with a bandwidth of Be was used after each Be EDFA Att3 BPF PIN detector. A pseudo random bit sequence with a word length 10 Balanced receiver of 2 -1 was used in this simulation. The launch power was set to Fig.1: Setup for receiver bandwidth optimisation study. 0dBm for SMF and –6dBm for DCF. 3. Results and Discussion Fig.2 shows the Q-value as a function of the optical and electrical filter bandwidths for all the formats studied, where the left plot for each format is the BTB case and the right one is after 8x80km transmission. For the MSK generation, we considered two approaches to achieve the sinusoidal weighting for offset quaternary PSK. The first approach used conventional carrier suppressed RZ (CSRZ) pulses generated by biasing a MZM (MZM1) at null point, which was driven by a clock signal with 2Vπ driving swing and a frequency of half bit rate [5]. While the second approach used an additional optical bandpass filter to remove the higher order side modes of the
a745_1.pdf CMJJ1.pdf
conventional CSRZ pulses and obtain pure dual-mode pulses. As shown in Fig. 2, for the BTB cases, the optimum optical and electrical filter bandwidths are around 3.3R and 0.8R for RZ-DPSK, and 3R and 0.8R for RZ-OOK respectively, which agree with the results obtained in [2, 4], where R refers to the data bit rate. However, the optimum optical and electrical filter bandwidths for MSK generated with conventional CSRZ pulses are 1.6R and 1.1R respectively. This indicates that the optimum optical filter bandwidth for this MSK signal is narrower than those for RZ-OOK and RZ-DPSK signals, and its optimum electrical filter bandwidth is higher than others. On the other hand, the optimum optical filter bandwidth for the MSK generated with pure dual-mode CSRZ pulses is broader, which is between 1.8R and 5R, and its optimum electrical filter bandwidth is around 0.8R. After 8x80km transmission, the optimum optical filter bandwidths for both MSK signals become very narrow, which are around 1.4R and 1.5R respectively. And the electrical filter bandwidths are around 1R and 0.75R respectively. However, the optimum filter bandwidths for the RZ-DPSK and RZ-OOK formats are just slightly reduced, which are around 2.9R and 2.6R respectively for optical filter, and 0.8R and 0.7R respectively for electrical filter. In general, the optimum optical and electrical filter bandwidths after transmission are narrower than those in the BTB cases, for all the formats, although the optimum range could be broader. This is because after transmission, the system OSNR is reduced, which is 35 dB under BTB case and 22 dB after transmission in this study. However, the filter bandwidth cannot be reduced unlimitedly, due to the trade off between the noise and inter-symbol interference (ISI) effect. It is also noted that the Q-value for both MSK signals after transmission does not reduce much. While, for the RZ-OOK, the Q-value penalty is around 1.6 after transmission. To investigate the impact of fiber nonlinear effect on the optimization of filter bandwidths, we ran another simulation by controlling the OSNR to be around 15dB for both the BTB and after transmission cases. The results showed if the OSNR remains the same, the optimum optical and electrical filter bandwidths remain almost the same between BTB and after transmission, for both the MSK formats. This observation tells that the optimum filter bandwidths are mainly determined by the OSNR in this evaluated system. 12 6.2
8 7
5.2
5.4
6 10
20
30
40
5 50
10
20
Bo (GHz)
30
6.2
8 6.2 6.0 5.8 10
40
50
Bo (GHz)
(c) RZ-DPSK (50% duty cycle)
10
20
20
6
5.2
30
40
5 50
Bo (GHz)
5.6 30
Bo (GHz)
5 40
50
10
20
4.6
7
6.2 5.8 5.6
4.4
7 4.4
6.0
4.2
5 30
40
50
9
Be (GHz)
30
10
Be (GHz)
20
Be (GHz)
Be (GHz)
10
7 5.4
9
7
5
8
5.8
6.0
5.6 9
7
6.0
5.6
(b) MSK using pure dual-mode pulses
9
5.4 5.2
9
5.4
6 5 50
40
30
7
10
5.2
Bo (GHz)
(a) MSK using conventional CSRZ pulses
5.6 5.8
20
Bo (GHz)
6.2
10 9
11
5.0
Be (GHz)
5.4
5.8 5.6
11
Be (GHz)
5.6
9
Be (GHz)
5.8
12
12 5.8 6.0
Be (GHz)
10
6.0
15 14 13 12 5.0 11 10 9 5.0 4.8 8 7 6 5 40 50 4.8
11
10
Bo (GHz)
5 20
30
40
50
Bo (GHz)
(d) RZ-OOK (50% duty cycle)
Fig. 2: Q-value as a function of optical and electrical filter bandwidths for different formats. (Left plot for each format is the BTB case, right one is after 8x80km transmission.)
4. Conclusion We studied the optimum receiver filter bandwidths for the MSK signals that we implemented recently. The MSK generated using conventional CSRZ pulses exhibits narrower optimal optical filter bandwidth while wider electrical filter bandwidth compared with RZ-OOK and RZ-DPSK. The MSK signal generated using pure dual-mode pulses has similar optimum range to RZ-OOK and RZ-DPSK at high OSNR, while with narrower optimal optical filter bandwidth at low OSNR. It is noted that the optimum range is mainly affected by OSNR in the evaluated system. 5. References 1. 2. 3. 4. 5.
P. J. Winzer, S. Pfenmigbauer, M. M. Strasser, and W. R. Leeb, JLT, vol. 19, no. 9, pp1263-1272, Sep. 2001. P. J. Winzer, S. Chandrasekher, and H. Kim, IEEE PTL, vol.15, no.6, pp840-842, June 2003. J. Rebola and A. V. T. Cartaxo, in Proc. LEOS 15th Annual meeting, pp780-781, 2002. L. Boivin and G. J. Pendock, in Proc. OAA 1999, pp292-295, 1999. J. Mo, Y. J. Wen, Y. Dong, Y. Wang, C. Lu, paper JTHB12, OFC2006.
