Optics Communications 255 (25) 41 45 www.elsevier.com/locate/optcom Effects of MPI noise on various modulation formats in distributed Raman amplified system S.B. Jun *, E.S. Son, H.Y. Choi, K.H. Han, Y.C. Chung Department of Electrical Engineering, Korea Advanced Institute of Science and Technology (KAIST), 373-1 Guseong-Dong, Yuseong-Gu, Daejeon 3571, Republic of Korea Received 3 February 25; received in revised form 15 May 25; accepted 31 May 25 Abstract We evaluated the effects of MPI noise on various modulation formats in a distributed Raman amplified system. The results show that is the most tolerant modulation format to MPI noise. Ó 25 Elsevier B.V. All rights reserved. PACS: 42.79.S Keywords: Multi-path interference; Modulation format; Raman amplifier 1. Introduction * Corresponding author. Tel.: +82 42 869 5456; fax: +82 42 869 341. E-mail addresses: freha@lsr.kaist.ac.kr (S.B. Jun), ychung@ ee.kaist.ac.kr (Y.C. Chung). Recently, there have been substantial interests in the distributed Raman amplifiers (DRAÕs) for the development of high-capacity WDM systems [1,2]. This is mainly because DRA can improve the optical signal-to-noise ratio and reduce the impact of fiber nonlinearity. However, as the gain of DRA increases, it could generate significant multipath interference (MPI) noise and limit the systemõs performance. Thus, for the optimization of DRA gain, it would be necessary to quantify the effect of MPI noise. Previously, there have been several reports on the effects of MPI noise in the distributed Raman amplified system [3 5]. These reports evaluated the effects of MPI noise in the presence of amplified spontaneous emission (ASE) noise on various modulation formats such as non-return-to-zero (N), return-to-zero (), differential phase-shift keying (), return-to-zero differential phase-shift keying (- ), return-to-zero alternate-mark-inversion (-AMI), and filtered phase-shaped binary transmission (PSBT). However, all these reports estimated the effects of MPI noise in a simulated 3-418/$ - see front matter Ó 25 Elsevier B.V. All rights reserved. doi:1.116/j.optcom.25.5.37
42 S.B. Jun et al. / Optics Communications 255 (25) 41 45 experimental condition (i.e., the MPI noise was not generated in a real Raman environment). This was mainly to have the capability of adjusting the relative amounts of MPI and ASE noises independently. In this paper, we investigated the effects of MPI noise on various modulation formats of 4-Gb/s signals (such as N,,, -, -AMI, and filtered PSBT) experimentally in a Raman amplified system. As a result, we could measure the system impairments caused by MPI and ASE noises while adjusting the Raman gain. The results show that the maximum improvement of Q-factor, achievable by using DRA, is different for various modulation formats by 1 db. 2. Experiments and results Fig. 1(a) shows the principle of measuring the MPI noise of modulated signal. To calibrate the MPI noise in the presence of other noises, we first generated the MPI noise by using a self-homodyne interferometer [6]. This optically generated MPI noise was measured in the frequency region below the lowest frequency component of the modulated signal (i.e., at the frequency region below 8 MHz for the 4 Gb/s signal with pattern length of 2 7 1) by using a RF spectrum analyzer. We then integrated the electrical spectrum in the range of 4 35 MHz, while varying the amount of the optically generated MPI noise. The background noises, caused by ASE noise, thermal noise, and shot noise, were subtracted from the integrated value. As a result, a linear relation between the measured value and the optically generated MPI noise was obtained, as shown in Fig. 1(b). Using this linear relation, we could estimate the MPI noise of the modulated signal in our experiment. Fig. 2 shows the experimental setup used to evaluate the effects of MPI and ASE noises on various modulation formats. We modulated a DFB laser, operating at 1559.3 nm, with 4-Gb/s signal obtained by electronically multiplexing four copies of 1-Gb/s signals. The pattern length of 1-Gb/s signal was set to be 2 7 1 for the MPI measurement, and 2 23 1 for the Q-factor measurement. The modulated signal was then sent to the transmission link consisted of 1-km long medium-dispersion fiber (a s =.22 db/km, a p =.26 db/km, D = 8.5 ps/km/ nm, A eff =71lm 2 ). The input signal power was set to be 2 dbm. This input power was low enough to neglect the deleterious nonlinear effects [7,8]. The fiber loss was compensated by using a Raman pump laser operating at 1465 nm. The maximum output power of this pump laser was 6 mw. For the measurement of MPI noise, we used an Erbium-doped fiber amplifier (EDFA) in conjunction -65 Signal Bit rate Noises measuring point a Noise frequency Noise Electrical Integrated b -7-75 -8-85 -9-95 -1 MPI Calibration Data Linear Fit -4-3 -2 MPI (db) Fig. 1. (a) Principle of measuring the MPI noise of the modulated signal and the measured electrical spectrum of MPI noise (inset). (b) Integrated electrical noise versus MPI noise.
S.B. Jun et al. / Optics Communications 255 (25) 41 45 43 4G Intensity 4G Phase 4G Phase 4G CLK N 1 MDF 4G CLK -AMI Filtered PSBT Delay-line Interferometer.3 nm Pump 1465 nm For MPI Measurement.5 nm RF PD Spectrum Analyzer OSA.5 nm Rx DCF Pre-Amp. For Q-factor Measurement Decision Circuit Fig. 2. Experimental setup to evaluate the effects of MPI and ASE noises on various modulation formats. with the Raman amplifier. This was necessary to maintain the optical power incident on the photodiode at a constant level while varying the Raman gain. The ASE ASE beat noise was suppressed by using an optical bandpass filter (bandwidth:.5 nm) placed in front of the photodiode. Under this condition, we measured the electrical spectra while varying the Raman gain for N signal, as shown in Fig. 3(a). These results were used together with the linear relation in Fig. 1(b) for the estimation of the MPI noise in the transmission link. Fig. 3(b) shows the optical signal-to-noise ratios measured against MPI and ASE noises as a function of the Raman gain. In this figure, OSN- R ASE and OSNR MPI represent the optical signalto-ase noise ratio and optical signal-to-mpi noise ratio, respectively. The measured data agreed well with the calculated values. Unlike ASE noise, MPI noise would have different effects for various modulation formats. This is because the spectral distributions of MPI and ASE noises are different from each other. While the MPI noise is a replica of the signal spectrum, the ASE noise is spread over the amplifierõs gainbandwidth. As a result, if the same receiver were used, the effect of ASE noise should be identical regardless of the modulation formats. However, in case of MPI noise, the effect could be different for various modulation formats due to their different optical bandwidths (i.e., when we used a modulation format with a broader optical bandwidth, a larger amount of MPI noise could result in the outside of the receiverõs bandwidth). To evaluate the effects of MPI noise on various modulation formats, we measured the Q-factors while varying the Raman gain. For this measurement, the Electrical -55-65 -7-75 Raman Gain db 5dB 1 db 15 db 2 db 22 db OSNR (db) 55 5 45 4 35 OSNR ASE OSNR MPI -8 1 2 3 3 2 4 6 8 1 12 14 16 18 2 22 a Frequency (MHz) b Raman Gain (db) Fig. 3. (a) Measured electrical spectra while varying the Raman gain. (b) The measured optical signal-to-noise ratios (for MPI or ASE noises) as a function of Raman gain. The solid and dashed lines are the calculated values limited by ASE and MPI noises, respectively.
44 S.B. Jun et al. / Optics Communications 255 (25) 41 45 Raman-amplified signal was passed through a dispersion-compensating fiber (DCF) module, and then sent to a pre-amplified receiver. The 3-dB bandwidth of this receiver was 32 GHz. Fig. 4 shows the Q-factor improvement achieved by Raman gain (db) Improvement Q-factor 7 6 5 4 3 2 1 Filtered PSBT N -AMI 2 4 6 8 1 12 14 16 18 2 22 Raman Gain (db) Fig. 4. Q-factor improvement achieved by Raman gain for various modulation formats such as N,,, -, -AMI, and filtered PSBT. The error bars were obtained from the 5 sets of independent measurements. The measured data agreed well with the calculated values (solid curves). for various modulation formats (such as N,,,, -AMI, and filtered PSBT). The baseline Q-factor was 15.56 db. The measured data agreed well with the calculated values. In this calculation, Q-factor pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi was obtained by using the equation, Q ¼ C= 1=OSNR ASE þ j=osnr MPI, where C is constant and j is the parameter indicating the relative impacts of ASE and MPI noises [4]. However, we noted that the parameter, j could be different for each modulation format. Thus, we measured the optical spectra of various modulation formats as shown in Fig. 5 and determined j by using the ratio of the optical power within the receiverõs bandwidth (i.e., shaded regions in Fig. 5) to the total optical power. As the Raman gain increased, the improvement of Q-factor achievable by DRA became different for various modulation formats by about 1 db. This was because the relative impact of MPI noise was increased with Raman gain, as expected from Fig. 3(b). The results also confirmed that formats (,, and -AMI) had more tolerance to MPI noise due to their wider signal bandwidths [3]. In particular, was the most tolerant modulation format to MPI noise. -1-2 N Filtered PSBT Optical -3-4 2xB e -1-2 -AMI Optical -3-4 Fig. 5. Measured optical spectra of various modulation formats (resolution BW:.2 nm). The shaded area indicates the optical power residing within the receiverõs bandwidth (32 GHz).
S.B. Jun et al. / Optics Communications 255 (25) 41 45 45 3. Summary We evaluated the effects of MPI noise on various modulation formats such as N,,, -, -AMI, and filter PSBT. Because of MPI noise, the Q-factor improvement achievable by DRA was different for various modulation formats by 1 db. The results show that was the most tolerant modulation format to MPI noise. References [1] H. Kidorf, K. Rottwitt, M. Nissov, M. Ma, E. Rabarijaona, IEEE Photon. Technol. Lett. 11 (May) (1999) 53. [2] D. Mahgerefteh, H.-Y. Yu, D.L. Butler, J. Goldhar, D. Wang, E. Golovchenko, A.N. Pilipetskii, C.R. Menyuk, L.J. Joneckis, Effect of randomly varying birefringence on the Raman gain in optical fibers, CLEOÕ97, paper CThW5, May 1997. [3] C.R.S. Fludger, Y. Zhu, V. Handerek, R.J. Mears, IEEE Electron. Lett. 37 (July) (21) 97. [4] J. Bromage, L.E. Nelson, C.H. Kim, P.J. Winzer, R.J. Essiambre, R.M. Jopson, Tech. Digest OFCÕ22 TuR3 (22). [5] C. Martinelli, G. Charlet, L. Pierre, J. Antona, D. Bayart, Tech. Digest OFCÕ23 FE3 (23). [6] C.R.S. Fludger, R.J. Mears, IEEE/OSA J. Lightwave Technol. 19 (Aug) (21) 536. [7] R.-J. Essiambre, P. Winzer, J. Bromage, Chul Han Kim, IEEE Photon. Technol. Lett 14 (July) (22) 914 916. [8] M. Daikoku, N. Yoshikane, I. Morita, Tech. Digest OFCÕ25 OFN2 (25).