You are on page 1of 3

IEEE PHOTONICS TECHNOLOGY LETTERS, VOL. 11, NO.

8, AUGUST 1999

991

NRZ Versus RZ in 1040-Gb/s Dispersion-Managed WDM Transmission Systems


M. I. Hayee, Member, IEEE, and A. E. Willner, Senior Member, IEEE
Abstract We compare nonreturn-to-zero (NRZ) with return-to-zero (RZ) modulation format for wavelength-divisionmultiplexed systems operating at data rates up to 40 Gb/s. We nd that in 1040-Gb/s dispersion-managed systems (single-mode ber alternating with dispersion compensating ber), NRZ is more adversely affected by nonlinearities, whereas RZ is more affected by dispersion. In this dispersion map, 10- and 20-Gb/s systems operate better using RZ modulation format because nonlinearity dominates. However, 40-Gb/s systems favor the usage of NRZ because dispersion becomes the key limiting factor at 40 Gb/s. Index TermsNRZ/RZ modulation, optical dispersion management, optical ber communication, optical ber nonlinearities, wavelength-division multiplexing.

OR wavelength-division-multiplexed (WDM) systems in which the data rate is 10 Gb/s/channel, the deleterious effects of dispersion and nonlinearity must be managed to achieve transmission over any appreciable distance [1]. Dispersion management, utilizing alternating ber segments of opposite dispersion values, is a key technique that keeps the total accumulated dispersion low while suppressing most nonlinear effects. In dispersion-managed systems utilizing single-mode ber (SMF) and dispersion compensating ber (DCF), the positive dispersion of SMF can be compensated by the large negative dispersion of DCF. In this scenario: 1) four-wave mixing (FWM) is signicantly reduced [1], [2] and 2) overall dispersion accumulation is minimized over a fairly wide wavelength range. However, self-phase modulation (SPM) and cross-phase modulation (XPM) still degrade the system performance. A recent study shows that in a 40-Gb/s single-channel system, return-to-zero (RZ) modulation is more optimal than NRZ in combating SPM [3] despite the fact that SPM is enhanced in RZ due to the higher pulse peak power [4]. Other reports show that RZ can take advantage of soliton-like pulse compression to perform better than NRZ for propagation in SMF [5], [6]. An issue still remains as to determining whether NRZ or RZ is the more optimal modulation scheme for use in WDM dispersion-managed systems. Intuitively, the following suppositions can be made: 1) due to higher peak power, RZ may suffer more nonlinearities, and due to shorter pulsewidth, RZ may suffer more dispersion and 2) due to a longer pulse
Manuscript received June 2, 1998; revised April 16, 1999. M. I. Hayee is presently with Tyco Submarine Systems, Eatontown, NJ 07724 USA. A. E. Willner is with the Department of Electrical Engineering-Systems, University of Southern California, Los Angeles, CA 90089-2565 USA. Publisher Item Identier S 1041-1135(99)05929-7.

time and longer interaction time between wavelengths, NRZ may suffer more nonlinearities. In this letter, we compare NRZ with RZ modulation for dispersion-managed systems (SMF alternating with dispersion compensating ber) operating at data rates up to 40 Gb/s. In 1040 Gb/s systems, we nd that NRZ is more adversely affected by nonlinearities whereas RZ is more affected by dispersion. Typically, 10- and 20Gb/s systems are limited mostly by nonlinearities, whereas 40 Gb/s systems are limited mostly by dispersion. Taken together, these two statements suggest that: 1) 10- and 20Gb/s systems, in general, operate better using RZ modulation because nonlinearity dominates and 2) 40-Gb/s systems, being limited mostly by dispersion, favor the usage of RZ for few channels but require NRZ as the number of channels increases. Our modeled system consists of alternate ber spans of 50 ps/nm km, and dispersion slope km of SMF ps/nm km) and 10 km of DCF ps/nm km, ps/nm km). Each ber span is and dispersion slope followed by an optical amplier with a noise gure of 6 dB to compensate for ber loss (0.2 dB/km for SMF and 0.5 dB/km for DCF). The launched power in SMF and DCF is same. The overall dispersion slope is 0.0125 ps/nm km. We have considered 16 WDM channels with uniform channel spacing (0.8 nm for 10- and 20-Gb/s systems, and 1.6 nm for 40Gb/s systems) with the center of the wavelength range being the point of complete dispersion compensation. A pseudorandom 64-bit data stream for each WDM channel is encoded using either nonreturn-to-zero (NRZ) or RZ (50% duty cycle) modulation. The transmitter chirp is assumed to be zero and and , bandwidth for NRZ and RZ is assumed to be is the bit rate. The combined electric respectively, where eld of all 16 channels is then propagated through the ber, and propagation of the resultant optical eld is simulated by solving the nonlinear Schr odinger equation using a split-step Fourier analysis [1]. The total eld approach is adopted for our simulation model, and the simulated spectral range is more than three times the bandwidth occupied by the WDM channels. Note that a variable step size is chosen for the simulations such that the nonlinear phase shift does not exceed 1 mrad/step, and the maximum step size is 100 m [1]. The affective areas of SMF and DCF are assumed to be 70 and 22 m , and the included nonlinear processes are SPM, XPM, FWM, and stimulated Raman scattering. At the receiver, each WDM channel is optically deand , respectively, multiplexed with a bandpass lter of for NRZ and RZ modulation formats. Each demultiplexed channel is then electrically low-pass ltered with a lter

10411135/99$10.00 1999 IEEE

992

IEEE PHOTONICS TECHNOLOGY LETTERS, VOL. 11, NO. 8, AUGUST 1999

Fig. 1. The factor of the worst of 16 channels versus average-power/channel for NRZ and RZ systems at (a) 10 Gb/s, (b) 20 Gb/s, and (c) 40 Gb/s.

Fig. 2. The factor for each of 16 channels modulated at (a) 10 Gb/s, (b) 20 Gb/s, and (c) 40 Gb/s. The average power/channel is 0.3 mW.

bandwidth . The system performance evaluation is based upon the eye closure penalty of each channel [1][3], [5]. To compare NRZ and RZ systems, we rst vary the average power per channel for a 16-channel system with and without including nonlinear effects; note that nonlinear effects are excluded from the simulation by making the nonlinearity coefcient zero. The resulting eye closure penalty of the poorest-performing channel is shown in Fig. 1(a)(c), respectively, for 10-, 20-, and 40-Gb/s systems. The poorest performing channel is channel #7 or Channel #8 in the cases of 10 and 20 Gb/s and is channel #1 or channel #16 in the case of 40 Gb/s. In 10- and 20-Gb/s systems, the eye closure penalty is large for lower channel powers because the EDFA noise is dominant [Fig. 1(a) and (b)]. For higher channel powers, the eye closure penalty decreases since the signal power now overcomes the EDFA noise. However, at the highest channel powers being considered, the eye closure penalty increases as a function of power due to nonlinearity. The penalty due to nonlinearity is more severe in NRZ as compared to RZ. Since RZ performs better than NRZ in both 10 and 20 Gb/s systems, we can conclude that RZ is less affected by nonlinearity than is NRZ. This is because: 1) isolated RZ pulses take advantage of soliton-like pulse compression in SMF, whereas the variable number of adjacent 1s in NRZ are nonuniformly degraded by nonlinearity, thereby causing the rail of 1s to spread [5], [7], and 2) long strings of 1s in NRZ as compared to RZ have a much longer cross-wavelength interaction time, thereby producing more severe penalty from nonlinearity [7], [8]. In a 40-Gb/s system [see Fig. 1(c)], the narrower RZ pulses are more susceptible to dispersion; note that dispersion-based

penalties grow inversely as the square of the pulsewidth [8]. Moreover, the required larger channel spacing of 1.6 nm for 40-Gb/s transmission produces more dispersion in the end channels since the total wavelength range is increased. An indication that RZ is more affected by dispersion can be derived from the fact that 40-Gb/s RZ systems have a larger eye closure penalty as compared to NRZ systems for low and moderate channel powers [Fig. 1(c)]. For higher channel powers, the penalty for NRZ increases because of nonlinearity, whereas no appreciable change in penalty is noticed for the RZ system. To isolate the degrading effects on individual channels, the penalties of all 16 channels are shown in Fig. 2(a)(c), respectively, for 10-, 20-, and 40-Gb/s systems. The penalties are shown with and without including nonlinear effects, and the average power per channel is 0.3 mW. This power value is chosen since ASE noise is overcome but nonlinear effects have not begun to dominate [Fig. 1(a) and (b)]. The penalty differential for all 16 channels at 10- and 20-Gb/s is negligible for NRZ and RZ when nonlinearities are not included. When nonlinearities are included, the penalty in NRZ increases more than RZ for all the channels. In a 16-channel 40-Gb/s system, the channel distribution in penalty is different for NRZ and RZ because dispersion dominates in RZ and affects the end channels more. In general, the nonlinearity induced penalty for all the channels in 10- and 20-Gb/s NRZ systems increases with the transmission distance. The eye closue penalty as a function of distance for the poorest-performing channel in a 16-channel system is shown in Fig. 3(a)(c), respectively for 10, 20,

HAYEE AND WILLNER: NRZ VERSUS RZ IN 1040-Gb/s DISPERSION-MANAGED WDM TRANSMISSION SYSTEMS

993

Fig. 3. The factor of the worst channel versus transmission distance for 16 channel NRZ and RZ systems at (a) 10 Gb/s, (b) 20 Gb/s, and (c) 40 Gb/s. The average power/channel is 0.3 mW.

Fig. 4. The factor versus dispersion slope of DCF for (a) 10 Gb/s, (b) 20 Gb/s, and (c) 40 Gb/s NRZ and RZ systems. The average power/channel is 0.3 mW and dispersion slope of SMF is 0.075 ps/nm2 1km.

and 40 Gb/s; note that the DCF length is not included in transmission distance. The penalty is larger for NRZ than for RZ at 10 and 20 Gb/s when nonlinearity is included. Furthermore, the penalty increases with distance in NRZ more rapidly as compared to RZ, thereby showing effectiveness of RZ modulation format in combatting nonlinearities. However, the penalty increases for 40 Gb/s more quickly with distance in RZ as compared to NRZ. Moreover, there is a negligible change in the penalty for RZ when nonlinearity is included, indicating that the penalty is mainly due to dispersion. For the 40-Gb/s NRZ system, the penalty increases slightly due to nonlinearities. The dispersion slope of DCF is an important parameter that inuences the wideband performance of dispersion-managed WDM systems. Therefore, we analyze in Fig. 4, the NRZ and RZ systems by varying the slope of dispersion for each 10-km span of DCF while keeping the slope of each 50-km span of SMF xed at 0.075 ps/nm km. The eye closure penalty of the poorest performing channel (Channel #1 or Channel #16) versus dispersion slope of DCF is shown in Fig. 4. When the combined accumulated dispersion slope has a small nonzero value (the regime where dispersion accumulation in all the channels is minimal and nonlinearity is the main degrading effect), RZ has a lower penalty than NRZ [Fig. 4(a)(c)] showing that RZ is less affected by nonlinearity than NRZ.

However, RZ systems have a smaller operational window in terms of dispersion slope of DCF as compared to NRZ [Fig. 4(a)(c)] because RZ modulation is more susceptible to dispersion due to high modulation bandwidth. REFERENCES
[1] R. W. Tkach, A. R. Chraplyvy, Fabrizio Forghieri, A. H. Gnauck, and R. M. Derosier, Four-photon mixing and high-speed WDM systems, J. Lightwave Technol., vol. 13, pp. 841849, May 1995. [2] D. Marcuse, A. R. Chraplyvy, and R. W. Tkach,Dependence of crossphase modulation on channel number in ber WDM systems, J. Lightwave Technol., vol. 12, pp. 885890, 1994. [3] D. Breuer and K. Petermann, Comparison of NRZ- and RZ-modulation format for 40-Gb/s TDM standard-ber systems, IEEE Photon. Technol. Lett., vol. 9, pp. 398400, Mar. 1997. [4] F. Forghieri, P. R. Prucnal, R. W. Tkach, and A. R. Chraplyvy, RZ versus NRZ in nonlinear WDM systems, IEEE Photon. Technol. Lett., vol. 9, pp. 10351037, July 1997. [5] K. Ennser and K. Petermann, Performance of RZ- versus NRZtransmission on standard single mode bers, IEEE Photon. Technol. Lett., vol. 8, pp. 443445, Mar. 1996. [6] R. M. Jopson, A. H. Gnauck, L. E. Nelson, L. D. Garrett, and C. Wolf, Evaluation of return to zero modulation for wavelengthdivision-multiplexed transmission over conventional single mode ber, presented at the Conf. Optical Fiber Communications 98, San Jose, CA, paper FE1. [7] L. Ding, E. A. Golovchenko, A. N. Pilipetskii, C. R. Menyuk, and P. K. A. Wai, Improvement of NRZ signal transmission through phase modulation, presented at the Conf. Optical Fiber Communications 97, Dallas, TX, paper TuJ. [8] I. P. Kaminow and T. L. Koch, Eds., Optical Fiber Telecommunications. New York: Academic, 1997, vol. IIIA.

You might also like