Você está na página 1de 5

2009 Second International Conference on Computer and Electrical Engineering

Spectral Analysis of Cross-Phase and Self-Phase Modulation for Different Compensation Techniques
Md. Shariful Islam , Arnab Dewanjee , Mehjabin Sultana Monjur Faculty, Dept. of EEE American International University - Bangladesh (AIUB) E-mail: shanto eee@yahoo.com aloukik1985@yahoo.com smehjabin@yahoo.com S. P. Majumder Professor, Dept. of EEE Bangladesh Univ. of Engg. & Tech. (BUET) Dhaka-1000, Bangladesh E-mail: spmajumder@eee.buet.ac.bd

AbstractThe spectral characteristics of cross-phase modulation (XPM) and self-phase modulation (SPM) in multi-span intensity-modulation direct-detection (IM-DD) optical systems are investigated. As constituent parameters, span-length and channel spacing, both are found to be strongly inuential regarding XPM and SPM. Short period dispersion managed bers (SPDMF), to suppress the XPM and SPM induced crosstalk, are found to serve the best in comparison with non-zero dispersion shifted ber (NZDSF) and also dispersion compensated ber (DCF). Keywords-WDM; XPM; SPM; SPDMF; NZDSF.

I. I NTRODUCTION Cross-phase modulation (XPM) and self-phase modulation (SPM) has an important impact on todays high speed dense wavelength division multiplexing (DWDM) optical ber communication systems [1], [2]. Though intensitymodulation direct-detection (IM-DD) optical systems are not particularly sensitive to phase modulation, due to chromatic dispersion of optical bers, phase modulation can be converted into intensity modulation [3] and, thus, degrades the system performance. Frequency response of XPM induced crosstalk in multispan WDM optical systems was analyzed previously. In this paper, we investigate the frequency response of both XPM and SPM induced crosstalk using the theoretical analysis previously introduced [4] for some comparative study for different compensation techniques and to investigate the qualitative effects of some link parameters on frequency response. Section II presents the necessary theoretical analysis to determine spectral characteristics of XPM and SPM induced crosstalk. In Section III we introduce a hypothetical WDM link to determine frequency response of XPM and SPM induced crosstalk on the basis of the theoretical analysis. A smooth high-pass characteristics of the XPM crosstalk is found considering a single span WDM system, while, in a multispan system, notches at some frequencies are evident, resulting from the interference between adjacent spans. The contribution of span length is comprehended which reveals
978-0-7695-3925-6/09 $26.00 2009 IEEE DOI 10.1109/ICCEE.2009.136 407

the detrimental effects of increasing length. This can be attributed to the gradual accumulation of XPM and SPM contamination through the ber length. Channel spacing, like length, is also found to be an important parameter deserving concern. A decrement in channel spacing between adjacent channels results in a deterioration of the XPM crosstalk scenario. However, the effect of increasing length is found to be more inuencing than that of altering channel spacing. A comparative analysis between XPM and SPM crosstalk is done which introduces SPM as a factor of no less concern. Dispersion compensation is considered as an effective measure to enervate the XPM and SPM crosstalk. In all the cases of our analysis we have investigated with different compensation techniques among which short period dispersion managed ber (SPDMF) is found to be the most promising one to replace the conventional NZDSF systems in order to suppress the two nonlinear crosstalk of our concern. II. T HEORETICAL A NALYSIS The theoretical analysis begins with the nonlinear wave propagation equation [5]. Let us consider the probe and pump optical signals, Aj (t, z ) and Ak (t, z ), copropagating in the same optical ber. The evolution of the probe wave is described by Aj (t, z ) 1 Aj (t, z ) = Aj (t, z ) z 2 vj t i2 2 Aj (t, z ) 2 t2 + ij pj (t, z )Aj (t, z ) + ij 2pk (t z/vk , z )Aj (t, z ) (1)

where is the attenuation coefcient of the ber, 2 is the ber chromatic dispersion parameter, j = 2n2 /(j Aef f ) is the nonlinear coefcient, n2 is the nonlinear refractive index, j and k are the probe and the pump signal wave2 lengths, Aef f is the ber effective core area, pk = |Ak | 2 and pj = |Aj | are optical powers of the pump and probe,

respectively. The last term in the right hand side(RHS) of (1) is slightly different from that in literature [5, Eq. 7.1.17]. Due to chromatic dispersion, the pump and the probe waves travel at different speed and this difference is taken into account in the calculation of XPM because it introduces walk-off between the two waves. vj and vk are used to represent the group velocities of probe and pump channel respectively. On the RHS of (1), the fourth term is responsible for SPM and the fth term is the XPM in the probe signal induced by the pump signal. The probe signal is considered to be operated in continuous wave (CW), whereas the pump signal is modulated with a sinusoidal wave at a frequency . The effect of SPM for both the probe and the pump channels are neglected in this XPM calculation. This approximation is valid as long as the pump signal waveform is not appreciably changed by the SPM induced distortion before its optical power is signicantly reduced by the ber attenuation. Under this approximation, the fourth term on the RHS of (1) is neglected in the XPM evaluation. XPM induced crosstalk involves the phase modulation generation through the nonlinear Kerr effect followed by the phase-noise to intensitynoise conversion through ber dispersion. The received power at the end of the ber contains XPM induced crosstalk as well as the attenuated signal power. Therefore, the probe output optical power with XPM-induced crosstalk in the time domain takes the form of [4] pjk (t, LN ) = pj (LN ) + Sjk (t, LN ) (2)

Figure 1.

A WDM point to point communication link

the probe and pump, respectively, and c is the velocity of light. For a dispersion slope ber, we can determine D() for a particular operating wavelength, , from its dispersion slope S0 and zero dispersion wavelength 0 , 4 using D() = (S0 /4)[1 (0 /) ] [6, chapter 6]. The basic difference between XPM and SPM is the channel spacing. By putting channel spacing jk = 0 in the equation of XPM and replacing the multiplication term 4 by 2, SPM induced electrical power spectral density, normalized to its power level, for an N -span system is found to be,
N

jj (, LN ) = S
i=1

2j pj (, 0) . sin[2
(n) (n) N ] n=i 2 L 2

(i)

(4)

where Sjk (t, LN ) is the XPM induced crosstalk power in time domain and pj (LN ) is the probe output at the end of the ber without XPM. Sjk (t, LN ) has a zero mean. The originally CW probe is intensity modulated by the pump through the XPM process. After a series of tedious derivations and approximations, the XPM induced electrical power spectral density in the probe channel, normalized to its power level, for an N -span system can be expressed as [4],
N

where, pj (, 0) is the power of the probe signal itself at the input of the ith span. III. R ESULTS AND D ISCUSSION At rst, we introduce a hypothetical point to point optical link. In our analysis we have simplied some parameters for ease of calculation. The erbium-doped ber amplier (EDFA) is supposed to give a at gain all over our concerned bandwidth. We also neglect any statistical variation of the ber parameters. For simplicity, we consider that the two copropagating waves have the same state of polarization and one way communication in this analysis. A hypothetical setup for our analysis is shown in Fig. 1. Two tunable lasers emitting at j , and k , respectively, are used as sources for the probe and the pump signals. The probe signal is CW while the pump signal is externally modulated. The two signals are combined by a 3-dB coupler and then sent to an EDFA to boost the power. A tunable optical lter is used at the receiver to select the probe signal and suppress the pump signal, the probe signal is detected by a photo diode. The parameters used for the analysis were chosen to be 0 = 1520.2 nm, S0 = .075 ps/km/nm2 , n2 = 2.351020 m2 /W, Aef f = 5.51011 m2 and = .25 dB/km. These values agree with nominal parameter values of the NZDSF practically used. Fig. 2 shows the normalized XPM frequency response measured at the end of a ber link consisting of a single
408

(i)

jk (, LN ) = S
i=1

4j pk (, 0)exp i . sin[2
N
(n) (n) N ] n=i 2 L (i) idjk 2

(i)

i 1

(n) d L(n) n=1 jk

(3)

(n) where, LN = is the total ber length in the n=1 L ( n ) ( n) system, L and 2 are ber length and dispersion of (i) the nth span [where L(0) = 0], pk (, 0) is the power ( n) of the pump wave at the input of the ith span and djk is the relative walk-off between two channels in the nth (0) span, [djk = 0]. Using a linear approximation, the walkoff djk can be expressed as djk = Djk , where D = (2c/2 )2 is the ber dispersion coefcient, jk and are, the wavelength spacing and average wavelength between

Figure 2. XPM frequency response in a system with single span (100 km) for NZDSF with different channel spacing ( = 0.08 dB/km)

Figure 4. XPM frequency response in a system with three spans (100 km each) for NZDSF with channel spacing 0.8 nm and = 0.25 dB/km

Figure 3. XPM frequency response in a system with single span (100 km and 300 km) for NZDSF with ch. spacing 0.8 nm and = 0.08 dB/km

100 km span of nonzero dispersion-shifted ber (NZDSF). The channel spacings used to obtain this gure were 0.2 nm (j = 1559 nm, k = 1559.2 nm), 0.8 nm (j = 1559 nm, k = 1559.8 nm) and 2 nm (j = 1559 nm, k = 1561 nm). High-pass characteristics are clearly demonstrated in each of the curves in Fig. 2. This is qualitatively different from the frequency dependence of phase-modulation obtained in [7]. In that analysis, the conversion from phase-modulation to intensity modulation through ber dispersion was not included. It is also observed that with the decrement of channel spacing the XPM power transfer increases. However, the plots for = .8 nm and .2 nm almost coincides. The difference of only approximately 5 dB is observed at the

very high modulation frequency. As for a practical optical pump signals the power related to such high frequency is relatively low, the effect of this variation is substantially suppressed. The effect of XPM by extending the link length without using any amplier is analyzed in Fig. 3. Since, extending the link without using any amplier is not that practical because of the high attenuation constant of normal bers, to examine the effect of length of ber a very low loss ber ( = .08 dB/km) is used. From Fig. 3, it is clear that by increasing the length of the link XPM induced crosstalk level is increased and a substantial difference of 20 dB is seen between these two curves. Thus XPM might set a constraint to setup long optical links without regeneration. However, bers with very low loss coefcient are still not commercially available. Therefore, to get a substantial amount of power at the receiver, we need to use internal ampliers such as EDFAs. Fig. 4 shows the plots of normalized power transfer function of an optical link with three spans (100 km each), each fed by an EDFA. Two plots are obtained for two different channel spacings, = .8 nm and 1.6 nm. Here we observe that the shape of XPM frequency response is dependent on the channel spacing in a system with multiple optical spans. The ripples in the XPM frequency response are due to the interference between XPM induced crosstalk which are created in different ber spans. It is observed that, the effect of a 300 km single span link is comparable to these 3 span 300 km link. A thorough speculation of Fig.2, Fig.3 and Fig.4 makes the fact clear that, the effect of increment of length of the optical link, irrespective of using inline amplier, is much severe than that of the decrement of channel spacing. Self Phase Modulation (SPM) is another nonlinear charac-

409

Figure 5. XPM vs SPM frequency response in a system with 1 span of 100 km for NZDSF with channel spacing 0.8 nm and = 0.25 dB/km

Figure 8. XPM frequency response in a system with two spans (100 km each) for different compensation techniques with channel spacing 0.8 nm (j = 1559 nm and k = 1559.8 nm) and = 0.25 dB/km

Figure 6.

Compensation with a dispersion compensating ber (DCF)

Figure 7.

Compensation with a short period dispersion managed ber

teristic that affects the system performance. A comparative analysis of XPM and SPM is done in Fig.5 using the same hypothetical optical link. It is observed that the effect of XPM and SPM is quite comparable. However, for a WDM system with more than two channels, the accumulated effect of XPM can be considerably larger than that of SPM. It is also observed that even for 2 spans, the spectral characteristic of SPM power transfer does not have any notch, unlike XPM. For the upcoming ultra high bitrate optical network, nonlinearities like XPM and SPM can be the most threatening factor that would limit the performance of an optical link. The active use of WDM systems reveals a new phenomenon: four wave mixing (FWM), which results in the degradation of system performance. Compensating the dispersion can be a good compensation technique for nonlinearities like XPM, SPM and FWM [6]. A technique to compensate

Figure 9. XPM frequency response in a system with one span (100 km) for different compensation techniques with channel spacing 0.8 nm (j = 1559 nm and k = 1559.8 nm) and = 0.25 dB/km

is to use dispersion compensating ber (DCF). The basic idea is simple: The positive dispersion stated above can be compensated by inserting a piece of single-mode ber with a negative dispersion characteristics so that the total dispersion of the link will be almost zero. Fig. 6 demonstrates the idea of DCF. It is recently reported that, short period dispersion managed ber (SPDMF) shown in Fig. 7 could suppress FWM efciently with minimal dispersion accumulation [8], [9]. Unlike the conventional dispersion managed bers, SPDMF could be accommodated easily within a single cable since it

410

Figure 10. SPM freq. response in a system with two spans (100 km each) for different compensation techniques with j = 1559 nm, = 0.25 dB/km

between XPM-induced crosstalk components created in different amplied ber spans has a strong impact on the overall frequency response of XPM crosstalk in the system. As a constituent parameter, channel spacing in a WDM system is found to be a very important one inuencing the XPM crosstalk, hence, limiting the maximum number of channels that can be accommodated into a single ber. Fiber length is reported to be another such important parameter. As XPM and SPM accumulate along the ber, an increment in length results in an increased noise. So, along with ber attenuation, XPM and SPM also add to set a limit to the maximum length of transmission. However, it is reported that the effect of increasing length in the presence of XPM can be more deleterious than that of the decreasing channel spacing. Dispersion compensation is shown to be an effective way to reduce XPM-induced crosstalk in IM-DD systems. Different schemes of dispersion compensation in multispan optical ber systems are evaluated and the use of SPDMF is found to be the most effective way to minimize the effect of both XPM and SPM induced crosstalk. R EFERENCES
[1] A.R. Chraplyvy, Limitations of lightwave technology imposed by optical ber nonlinearities, J. Lightwave Technol., vol 8, pp. 1548-1557, 1999. [2] D. Marcuse, A. R. Chraplyvy and R. W. Tkach, Dependence of cross-phase modulation on channel number in ber wdm system, J. of Lightwave Technolo., vol. 12, no. 5, May 1994. [3] J. Wang and K. Petemm , Small signal analysis for dispersive Optical ber communication systems, J. Lightwave Technol. Lett, vol. 10, pp. 96-100. 1992. [4] Rongqing Hui, Kenneth R. Demarest and Christopher T. Allen, Cross Phase Modulation in Multispan WDM Optical Fiber Systems, J. Lightwave Technol., vol. 17, no. 6,June 1999. [5] Govind P. Agrawal. Nonlinear Fiber Optics, Third edition. Academic Press, An Imprint of Elsevier, 2001. [6] Djafar K. Mynbaev and Lowell L. Scheiner, Fiber-Optic Communication Technology, Pearson Education, Inc, 2007. [7] T.K. Chiang, M. E. Marhic N. Kagi, and L. Kazovsky. Crossphase modulation in ber links with multiple optical ampliers and dispersion compensators, J. Lightwave Technol., vol. 14, pp. 249-260, Santa Barbara, California, USA, 1996. [8] H. S. Chung, H. Kim, S. E. Jin, E. S. Son, D. W. Kim, K. M. Lee, H. Y. Park, and Y. C. Chung, 320 Gb/s(32 x 10 Gb/s) wdm transmission with 50GHz channel spacing over 564 km of short-period dispersion-managed ber (perfect cable), IEEE Photonic Technolo. Lett., vol. 12, pp. 1397-1399, Oct, 2000. [9] H. S. Chung, H. Kim, S. E. Jin, E. S. Son, D. W. Kim, K. M. Lee, H. Y. Park, and Y. C. Chung., 640 Gb/s(32 x 20 Gb/s) wdm transmission with .4(bit/s)hz spectral efciency using short-period dispersion-managed ber (perfect cable). In OFC Tech. Dig., Anaheim, CA, 2001, Paper ThF6.

is composed of short positive and negative dispersion bers. There are two different ber sections in SPDMF, one with positive dispersion D1 , and the other with negative dispersion D2 . Two successive sections constitute one dispersion uctuation period, referred to as a segment, with positive dispersion ber length l1 and negative dispersion ber length l2 , respectively. From the analysis of Fig. 8 it is evident that, XPM crosstalk level is the highest for NZDSF. DCF gives better result but the improvement is not that promising. However, the use of SPDMF seems to be promising where suppressing XPM crosstalk is a major concern. Fig. 8 shows that with l1 = l2 = 5 km and l1 = l2 = 10 km, with negative dispersion coefcient -15 ps/nm.km, the difference between the PSDs is insignicant. Again, from Fig. 9 we nd that when negative dispersion equals -18 ps/nm.km, that is complete compensation, gives the best result. However, it has been reported that the dispersion of the negative section should be larger than -15 ps/nm.km to avoid the deleterious effect of FWM [8]. Thus for the simultaneous suppression of XPM and FWM, it would be desirable to set the dispersion of the negative section to be about -14 to -15 ps/nm.km. Fig. 10, which plots the spectral density of SPM induced intensity uctuation, extracts the similar results compared to that of XPM using different compensation techniques. From the above analysis we can conclude that, using SPDMF gives the optimum result for suppressing the nonlinear effects like XPM and SPM considering FWM. IV. C ONCLUSION We have investigated the spectral characteristics of XPM and SPM in multispan IM-DD optical system. Interference

411

Você também pode gostar