CHAPTER 3 PERFORMANCE OF MODULATION FORMATS ON DWDM OPTICAL SYSTEMS

67 CHAPTER 3 PERFORMANCE OF MODULATION FORMATS ON DWDM OPTICAL SYSTEMS 3.1 INTRODUCTION The need for higher transmission rate in Dense Wavelength Division optical systems necessitates the selection of a suitable modulation format for the efficiency of the optical system and has been a key issue in recent research (Vassilieva et al 2001). Generally modulation formats are classified into On Off Keying (OOK) and Phase Shift Keying (PSK) techniques (Idler et al 2003). Over the last few years, novel modulation formats with improved performance than the NRZ scheme have been suggested and investigated, (Matsuda et al 1998). From the literature it is found that by adapting a Return-to-Zero (RZ) format we can improve the receiver sensitivity and non-linear tolerance (Winzer and Kalmar 1999, Caspar et al 1999), but at the extra cost of one additional modulator and drive circuitry in the transmitter. Phase modulation combined with a balanced receiver offers a very attractive 3 db improved receiver sensitivity compared to OOK, (Ferber et al 2003). Recently, many OOK formats with additional phase modulation have been shown to perform very well under certain circumstances, for example Chirped-RZ scheme (Bergano et al 1998) which however, adds further complexity to the transmitter. Four-level phase modulation - Differential

68 Quadrature Phase Shift Keying (DQPSK) has also been studied recently (Griffin et al 2003, Wree et al 2002, Cho et al 2003). In this chapter the impact of the fiber non-linear effect is studied for a 32 channel DWDM system for RZ and NRZ modulation formats using the OPT package and the spectrum of Duobinary and CSRZ modulation formats are also observed. The spectrum is also observed for a single channel system as by, (Binh and Csematony 2003) and a sixteen channel system for RZ and NRZ modulation formats with phase shift keying by carrying out a simulation using MATLAB simulink for our fiber and data rate specifications. The multi channel CSRZ DQPSK modulation format for 16 Unequal spacing channels is also simulated and the Q factor is observed. In addition, the impact of filtering techniques and dispersion compensating techniques are also analyzed in terms of Q factor and Eye Opening Penalty to identify methods for improving the spectral efficiency. 3.2 IMPACT OF NON-LINEAR EFFECTS ON MODULATION FORMATS 3.2.1 RZ, NRZ and CSRZ All modulation formats can be divided into NRZ-based and RZbased modulation formats, (Mohs et al 2000). In this thesis, we have made an effort to study these modulation formats on a 32 channel DWDM system at 40 Gb/s bit rate per channel accounting for the impact of non-linear effects like XPM and FWM. A simulation scenario is set up initially as shown in Figure 3.1, (Hodzic 2004) and applied in this work for our specifications. The system and fiber parameters used for the simulation are listed in Tables 3.1 and 3.2

69 respectively. The channels allotted are chosen between 191.9 and 195 THz with 100 GHz spacing in the C band, with EDFA based amplification and dispersion compensating fibers. The transmission data are coded into NRZ sequences and RZ sequences and the corresponding Q factors are measured and plotted with and without the inclusion of non-linear effects. The Q factors obtained with and without the non-linear effects are plotted with respect to the channel numbers as shown in Figure 3.2. The Quality Factor in this case is defined as the center frequency divided by the signal bandwidth in Hertz. The non-linearity comes into picture when the signal power per channel is higher, in this case 6 mws. When the power is reduced to 2 mws the non-linear effects observed are negligible. Figure 3.1 Simulation setup for 32 channel DWDM system

70 Table 3.1 Simulation parameters for the system at 40 Gb/s Parameter Rise time of electrical signal MZM extinction ratio Duo binary low pass filter Duo binary low pass filter cut off frequency Laser frequencies Laser Line width Simulation bits PRBS length NLSE step size Sample rate EDFA noise figure Splice loss Optical filter Optical filter 3 db bandwidth Values 5 ps Infinite Fourth order Bessel type 11.7 GHz 193.1 THz - 195 THz 10 Hz 1024 2 10 1 1000 m (or) 0.06 rad 32 samples / bit 5 db 0 db Second Order Super Gaussian 100.0 GHz Table 3.2 Fiber parameters used in the simulation for 32 channels, Keiser(2000) Parameters SMF DCF Length 80 km Variable ( 14.5 kms) Attenuation 0.2 db/km 0.25 db/km GVD 16 ps/nm/km -72 ps/nm/km Slope 0.08 ps/nm 2 0.08 ps/nm 2 Effective Area 80 µm 2 30 µm 2

71 Figure 3.2 Q-factor performance of NRZ modulation format It is observed from Figure 3.2 that the Q factor is degraded with the inclusion of non-linearities. It is also noted that the maximum Q factor degradation is seen for the center channel. When the impact of non-linearity is not included the center channel has a Q factor of 18 which gets reduced to 13 in the presence of non-linear effects. Hence we come to the conclusion that non-linear effects will affect the spectral efficiency and the performance degradation is maximum for the channels placed in the center. The Q factor obtained with and without the non-linear effects in this case of RZ shaping, is shown in Figure 3.3. In the case of RZ format, the width of the optical signal is smaller than its bit period. It is observed from the figure that there is degradation in Q factor due to non-linearity, however the Q factor obtained is much higher than the case of NRZ formatting. The reason for its superior performance is probably due to its return-to-zero characteristic.

72 Figure 3.3 Q-factor performance of RZ modulation format Duo binary format whose spectral width is half of the standard NRZ format is a possible way to increase spectral efficiency but it s resistance to nonlinear effects is not so large because the phase information is vulnerable to interaction with Amplified spontaneous emission noise and SPM (Yonennaga et al 1995, Keiser et al 2000). Thus this type is suitable only for short distance DWDM systems. The spectrum of the Duo binary signaling scheme obtained with our simulation for the design parameters from Table 3.1 is shown in Figure 3.4. The narrow pulse nature of RZ format has a wider spectrum leading to less spectrum efficiency in a DWDM system. To overcome this difficulty and to improve the spectrum efficiency Carrier-suppressed RZ (CSRZ) modulation has been recently proposed for high bit rate transmission

73 Figure 3.4 Spectrum of the duo binary signal systems, and has been intensively investigated in numerical and experimental works (Miyamoto et al 2000). The transmitter and the fiber section for the CSRZ format is shown in Figures 3.5 and 3.6, respectively as in Bosco et al (2002). The spectrum obtained is shown in Figure 3.7. The carrier component of the CSRZ signal spectrum is suppressed due to the external modulation at zero point in the second MZM. The CSRZ pulses possess a RZ signal shape with an optical phase difference of π between adjacent bits. We have carried out investigation with the CSRZ modulation format along with the phase modulations in this thesis. The spectrum of the CSRZ modulation has been studied in this work for our specification in the Table 3.1.

74 20 G clock CW laser MZM1 MZM2 40 Gbs 20 G clock Figure 3.5 Transmitter section of CSRZ format EDFA 1 EDFA 2 EDFA 3 SMF SMF SMF Loop control Number = 6 Figure 3.6 Fiber section of CSRZ format

75 Figure 3.7 Spectrum of CSRZ format 3.2.2 Optimal Pre-Transmission Filter for DWDM Systems A spectral efficiency of 0.1 b/s/hz is realized with 100 channels of 100 GHz spacing and 10 Gb/s rate per channel. Spectral efficiency is generally defined as the ratio of the per channel data rate to the channel spacing in WDM systems. One way to improve the spectral efficiency is to increase the data rate from 10 to 40 Gb/s and maintaining an optimum channel spacing ( Ito et al 2000). The minimum channel spacing is limited by the non-linear effects and the data rate efficiency depends on the modulation format. Due to reduced spectral width, CSRZ modulation shows an increased dispersion tolerance and is also more robust to non-linear impairments than conventional RZ and NRZ systems, (Miyamoto et al 1999).

76 The non-linear tolerance of CSRZ modulation can be enhanced by the implementation of pre-chirp at the transmitter side (Sano and Miyamoto 2001) but the amount of pre-chirp has to be carefully optimized in order to avoid the increase of linear cross talk and waveform distortions. The robustness of CSRZ modulation to narrow band filtering can be improved and hence can be beneficial for DWDM systems (Morita and Edagawa 2003). In the reported literature, various transmission filters like Butterworth, Chebychev, Dielectric, Fiber Bragg Grating, Arrayed Waveguide Grating and Mech-Zender Interferometer Filter have been analyzed (Hodzic et al 2003). In this section, the performances of a Gaussian Filter, a Second order Super Gaussian Filter and a Flat top Bragg Grating Filter are compared by simulating the setup shown in Figure 3.8. Narrow band filters with sharp filter edges and flat pass band represent the optimal solution for DWDM systems Teraxion (2002). For the numerical investigation presented here, the characteristics of flat top filters are emulated using Gaussian filters of higher order degree n 2.In this work, we have measured the performance of the filter by finding the Q factor.the frequency response of such a Super Gaussian filter is defined by the transfer function, Hodzic (2004), 2n ( ln 2 2(f fc) T(f ) exp f 3dB (3.1) where f c is the filter central frequency, n defines the order of the filter and Δf 3db represents the 3-dB optical bandwidth of the filter. The increase in filter order results in increased steepness of the filter edges. In this section, we have used a second order Super Gaussian filter. A binary sequence of 2 9 bits is considered to assess the performance of a single channel case. As these filters have a real transfer function, delay distortion does not occur. Using low dispersion fibers reduces the problem of dispersion, but increases the effect of

77 FWM and using high dispersion fibers results in decreasing the effects of XPM, (Kai Song 2000). The simulation parameters used are listed Table 3.3. Figure 3.9 shows the eye diagram with measured Q factor in our simulation using OPT simulation package. Tx Filter Fiber EDFA Detector Figure 3.8 Simulation set up for filter performance study Table 3.3 Simulation parameters used in the setup to study filter performance S.No. Parameter Values 1 Number of loops 09 2. Input power per channel 0.25 mw 3. SMF and DCF lengths 80 kms and 14.5 kms per span 4. EDFA parameters Noise Figure 4 db, Gain 30 db, two per span 5. WDM channels, data rate 192.5-193.7 THz, 40 Gb/s 6. Transmitter filter bandwidth 80 GHz bandwidth Figure 3.9 Q factor measured at the receiver output

78 Figure 3.10 shows the Q factor measured for our specification and simulation parameters at the receiver output for varying fiber lengths for various filters like Bragg Grating, Gaussian, Super Gaussian and Fabry Perot Filters. The reduction in Q factor for increasing distances indicates the necessity for a dispersion management technique. Super Gaussian filter of higher order, in this case four, is observed to give the highest Q factor. Bragg grating Filter shows a performance close to that of the Super Gaussian filter. Super Gaussian filter show better tolerance to dispersion and nonlinearity due to it s reduced spectral width as the order increases. Hence more number of channels could be incorporated in the available spectral width of the optical channel resulting in an improvement in the spectral efficiency. This filter has no negative side lobes hence this can be incorporated for the dispersion managing schemes to limit the interactions of XPM and FWM effects. Figure 3.10 Receiver output Q factor variation with distance

79 3.3 MODELING THE PHASE SHIFT KEYING METHODS Phase Shift Keying (PSK) uses the phase of the optical carrier to encode information. The Differential PSK has larger resistance to non linear effects at higher data rates, (Rhode et al 2000).Wree et al (2002) investigated the improvement in the spectral efficiency for RZ DQPSK using balanced detection. The Differential Quadrature Phase Shift Keying (DQPSK) modulation characterized by a symbol rate that is only half of the nominal bit rate, results in better narrow-band filtering characteristics and an improved dispersion and PMD tolerance (Hoshida et al 2003). The interaction between the ASE noise and the non-linearities will modify the probability density function used for BER calculation. It has been shown that (Ho 2004), nonlinear noise induced by ASE does not have the characteristics of a Gaussian pdf. The general architecture used for system modeling is shown in Figure 3.11, (Binh and cheung 2005). The pseudo random data bit sequence is simulated using the Bernoulli binary generator available in the communication block set of the SIMULINK. CW LASER is modeled by the sine wave generator available in the signal processing block set. The main section for simulating NRZ/RZ-DPSK modulation techniques is shown in Figure 3.12, consisting of DPSK block and the Phase Modulation (PM) block, ( Liem et al 2005). The NRZ electrical signal is first encoded by DPSK encoder and the encoded electrical signal is then used to drive an electro-optic phase modulator to generate a DPSK optical signal. The expanded simulation block for DPSK and PM are shown in Figures 3.13 and 3.14, respectively, ( Liem et al 2005). Inside the PM block, the input port 1 is

80 fed with the optical carrier and input port 2 is fed with the DPSK electrical data. To produce the NRZ-DPSK optical signal, the complex phase shift block is then used to phase shift the optical phase by π when the data bit is 1 and a phase shift of zero when the data bit is 0. In the case of RZ DPSK an additional Intensity modulator is added and in this additional MZIM, the NRZ-DPSK optical signal from output of the PM block will be sampled by a pulse generator to achieve the desired RZ-DPSK signal. The sampling pulse train is synchronized with the input electrical sequence and is at the same rate as the data rate. The results are observed and studied for the simulink models taken from the Binh etal 2006 for our date rate and specifications towards the spectral improvement. In the receiver block, the NRZ-DPSK optical signal is demodulated and passed through a low pass filter to remove the carrier. A one-bit-delay Mach-Zehnder Interferometer (MZI) is usually used as a DPSK optical receiver as shown in Figure 3.15, ( Binh et al 2006). In a DPSK balanced receiver, a photodiode is used at each MZI output and then the two photocurrents are combined (logical subtract) to double the signal level (Binh et al 2005). The spectrum of the NRZ-DPSK and RZ-DPSK modulated signals obtained from simulation are shown in Figures 3.16 and 3.17, respectively for our specification. Based on the investigation made by the models proposed by Binh et al monash university we have simulated the RZ and NRZ DPSK spectrum for our data rate and specifications as discussed.

Input Output Encoder signal shaping circuit Decoder Demodulator Modulator Optical fiber Amplifier Optical source Optical fiber EDFA Optical detector Figure 3.11 General architecture for system modeling (Binh et al 2006) 81

Figure 3.12 Main section for simulation of NRZ/RZ-DPSK modulation techniques 82

Figure 3.13 Expansion of DPSK block 83

Figure 3.14 Phase modulation block 84

Figure 3.15 Receiver block 85

86 Frequency (GHz) Figure 3.16 NRZ -DPSK spectrum at fiber input Frequency (GHz) Figure 3.17 Spectrum of the RZ-DPSK

87 An optical transmission with very high spectral efficiency is one of the objectives of this research work. Various modulation formats for very high spectral efficiency have been investigated in the literature, such as Duobinary (Yonenaga and Kuwano 1997), NRZ VSB (Idler et al 2002) and Narrow Band RZ (Gnauck et al 2003) to name a few. A common feature of all these techniques is that of narrow signal bandwidth with a reduction in the symbol rate. DQPSK is another spectrally efficient modulation format by which two bits per symbol is transmitted. DQPSK is a four-symbol format equivalent to phases of {0, π/2, π, 3π/2}. Depending on the desired di-bit combination to be encoded, the difference in phase, Δφ, between the two adjacent symbols (optical carrier pulses) is varied systematically, (Griffin and carter 2002). The signal spectra before and after a fiber length of 450 Kms for RZ-DQPSK are as depicted in Figures 3.18 and 3.19, respectively. The general simulation block, received signal eye-diagram and signal spectra for CSRZ-DQPSK are shown in Figures 3.20,3.21 and 3.22, respectively. In order to built the simulation block in simulink for CSRZ DQPSK we have studied the models and lay out for NRZ / RZ DPSK and RZ DQPSK from ( Binh and Laville 2005) series of Monash university and the results which are discussed in this chapter for the above formats are obtained for our model parameters.hence in our work the model of CSRZ DQPSK has been developed based on the other modulation formats simulation, the results are obtained for the CSRZ DQPSK spectrum and eye pattern at the detector output (Ramprasad and Meenakshi 2005).

88 Figure 3.18 RZ-DQPSK spectrum before the fiber Figure 3.19 RZ-DQPSK after propagating 450 kms in the fiber

Figure 3.20 CSRZ-DQPSK simulation set up 89

90 Amplitude Time (bit period) Figure 3.21 Receiver eye pattern CSRZ-DQPSK Figure 3.22 Spectrum of CSRZ-DQPSK

91 3.4 EYE OPENING PENALTY A broadened eye opening is seen at the receiver for ideal optical transmission. The comparison of the different modulation formats is done with reference to the Eye Opening Penalty (EOP) which is defined as the ratio of the difference in the eye openings of the mark and space state to the difference in their corresponding variance. In order to compare the transmission characteristics of PSK-based modulation formats, a 40 Gb/s DWDM transmission system with Unequally Spaced Channels and 7 spans of SSMF fibers of span length 80 kms is considered. The EOP values are measured from the simulation model shown in Figure 3.23 using MATLAB SIMULINK by varying the transmission distance for 16 channels with USC scheme. The transmitter, receiver and fiber parameters used are as listed in Tables 3.4 and 3.5. Figure 3.24 provides a comparison of the EOP at different fiber spans for the NRZ-DPSK, RZ-DPSK and the CSRZ-DQPSK modulation formats. The figure shows that CSRZ-DQPSK offers the best transmission results up to 560 kms. This is certainly due to its low symbol rate, which suppresses the effects of fiber dispersion and subsequent nonlinear effects on the signal as well as good narrow-band filtering characteristics. However when compared to RZ-DPSK, the signal quality steadily decreases at greater transmission spans. This can be attributed to the effects of ASE induced phase noise, which accumulate as the transmission distance increases, resulting in

Figure 3.23 Simulation model for measuring EOP and Q factor 92

Table 3.4 Simulation parameters used to determine Q factor for USC and ESC scheme. S.No. Parameter Values used in the model 1. PRBS 2.5 Gb/s,2 23-1,10Gb/s 2. Pulse generator NRZ /RZ Encoding 3. Laser Diode 3 mw power per channel line width 10 MHz 4. Modulator MUX Mech zender modulator Gaussian MUX filter at the transmitter 5. EDFA Gain 12 db, Noise factor 4 db. 6. Fiber Nonlinear dispersive fiber length 60 km, attenuation 0.22db/km 7. Optical filter Trapezoidal filter with Zero db bandwidth 45 GHz, Cut off Bandwidth 50 GHz, Cut off Magnitude 30 db. 8. Photo detector Responsivity 1 A/W,Dark current 10 na 9. Low pass Filter Bessel filter Bandwidth 1.875G Hz 10. Power per channel Variable power in milli watts per channel 93

94 Table 3.5 Non-linear dispersive fiber parameters used in the simulation Lucent (2001) Parameters Values α attenuation 0.25 db/km Input coupling efficiency -1 db Output coupling efficiency 0.022dB GVD constant 4.5 ps/nm/km Dispersion slope constant 0.11ps/nm 2 /km Effective Area 72 m 2 N2 constant 2.6 e-20 m 2 /w Peak Raman gain coefficient 9.9 e-014 m/w Pump wavelength 1000 nm Raman self shift time 5 fsec Figure 3.24 EOP measurements for the modulation formats

95 larger phase fluctuations and possible errors in the detection process. RZ-DPSK can cope better with these effects and is less severely affected over the entire transmission span, because of better error tolerance at the receiver compared to DQPSK. NRZ-DPSK has higher EOP values and larger variations in the EOP due to its reduced non-linear tolerance, (Ramprasad and Meenakshi 2005 ). 3.5 IMPACT OF FWM ON MODULATION FORMATS 3.5.1 Q Factor Measurement Based on the simulation model, the Q factor is measured for various modulation formats namely RZ, NRZ, CSRZ and VSB-RZ. The block diagram with optimum filter design to generate VSB-RZ and CSRZ are obtained from the previous section. The fiber is modeled as non-linear and dispersive thereby giving rise to FWM components in the DWDM system considered for simulation. Table 3.6 shows the measured Q factor for NRZ modulation with four channels at 10 Gb/s and other parameters same as shown in Tables 3.4 and 3.5. The channel allocations using ESC technique and orthogonal coding based USC techniques, for four channels are {193.1, 193.2, 193.3, 193.4} THz and {193.1, 193.3, 193.4, 193.8} THz respectively. The corresponding results obtained for RZ categories are tabulated in Tables 3.7, 3.8 and 3.9. It is found from the Tables 3.6, 3.7, 3.8 and 3.9 that the Q factor for both ESC and USC schemes decreases in spite of an increase in the channel power of the system for all the modulation formats. This is due to the presence of Four wave mixing effect. In general RZ format shows better Q factor values and hence is more non-linear tolerant. The non-linear tolerant

96 Table 3.6 Q factor measured for various input powers for ESC and USC schemes under NRZ modulation - 4 channel DWDM Power in mw 0.25 0.5 0.75 1.0 1.25 1.5 Q factor ( ESC ) 14.04 6.977 4.03 2.853 1.839 1.222 Q factor ( USC ) 16.97 9.553 6.904 4.982 3.079 2.998 Table 3.7 Q factor measured for various input powers for ESC and USC schemes under RZ modulation - 4 channel DWDM Power in mw 0.25 0.5 0.75 1.0 1.25 1.5 Q factor ( ESC ) 28.02 12.89 8.99 5.12 3.89 2.44 Q factor ( USC ) 32.24 18.78 13.98 9.88 6.98 6.99 Table 3.8 Q factor measured for various input powers for ESC and USC schemes under CSRZ modulation - 4 channel DWDM Power in mw 0.25 0.5 0.75 1.0 1.25 1.5 Q factor ( ESC ) 32.06 16.01 12.98 9.66 7.83 6.89 Q factor ( USC ) 38.27 24.53 16.24 15.45 12.37 12.23 Table 3.9 Q factor measured for various input powers for ESC and USC schemes under VSB-RZ modulation - 4 channel DWDM Power in mw 0.25 0.5 0.75 1.0 1.25 1.5 Q factor ( ESC ) 33.85 17.55 14.56 11.66 9.39 8.222 Q factor ( USC ) 41.97 26.65 18.904 17.982 14.079 14.998

97 characteristics is combined with the filtering effect at the transmitter side to get vestigial side band RZ format which shows good overall performance in the presence of FWM. Comparing Q factor for ESC and USC schemes under constant input powers; it is observed that the Q factor values are higher for USC than ESC irrespective of the modulation format implying that USC schemes are more FWM tolerant, (Ram prasad and Meenakshi 2006). Figures 3.25 and 3.26 show the spectrum of the 4 channel ESC and USC schemes. The strength or the power of the inter modulation FWM cross talk products falling on the desired channels for both ESC and USC scheme are shown. It is seen that many cross talk inter modulation products are present in the desired band of the four channels for ESC and lesser inter modulation products are present in the desired channel band as cross talk components for the USC scheme. Figure 3.25 Output spectrum of four channels in ESC

98 Figure 3.26 Output spectrum of four channels in USC 3.5.2 Simulation Results for 16 Channels The simulation is carried out for 16 DWDM channels with the same system parameters shown in Tables 3.4 and 3.5 but the data rate is chosen as 40 Gb/s per channel. The results obtained for NRZ, RZ, CS-RZ and VSB-RZ are tabulated as shown in Tables 3.10, 3.11, 3.12 and 3.13 respectively. Table 3.10 Q factor measured for various input powers for ESC and USC Schemes under NRZ modulation - 16 channel DWDM Power in mw 0.25 0.5 0.75 1.0 1.25 1.5 Q factor ( ESC ) 11.09 4.977 2.03 1.84 1.2 1.0 Q factor ( USC ) 14.34 7.78 4.98 4.45 4.34 3.45

99 Table 3.11 Q factor measured for various input powers for ESC and USC Schemes under RZ modulation - 16 channel DWDM Power in mw 0.25 0.5 0.75 1.0 1.25 1.5 Q factor ( ESC ) 26.23 11.45 06.56 3.67 3.78 2.87 Q factor ( USC ) 36.26 16.78 11.57 07.76 04.76 04.56 Table 3.12 Q factor measured for various input powers for ESC and USC Schemes under CSRZ modulation - 16 channel DWDM Power in mw 0.25 0.5 0.75 1.0 1.25 1.5 Q factor ( ESC ) 31.67 15.78 11.64 7.546 5.76 4.676 Q factor ( USC ) 36.454 22.656 14.75 13.565 10.75 10.87 Table 3.13 Q factor measured for various input powers for ESC and USC Schemes under VSB-RZ modulation - 16 channel DWDM Power in mw 0.25 0.5 0.75 1.0 1.25 1.5 Q factor ( ESC ) 31.66 15.45 12.77 9.46 7.77 6.74 Q factor ( USC ) 37.45 24.36 16.26 15.47 12.57 12.11 On observing the tabulated values for 16 channel system, it is observed that the Q factor values for USC are reasonable even at higher channel powers and is almost double that of ESC schemes. Hence we conclude that, better suppression of Four wave mixing effect at higher data rate is achieved by the Unequal Spacing Channel allocation using orthogonal optical codes. Further, the comparison of Q factor values suggests that a

100 combination of orthogonal code based USC and VSB-RZ formatting can effectively combat FWM even at higher channel powers. Figures 3.27 and 3.28 show the received Eye diagram for NRZ and RZ coded 16 channel DWDM system under ESC and USC schemes, respectively. The eye is seen to be more distinct for RZ under USC compared to that of NRZ under ESC. Figure 3.27 Measured Q factor and Eye diagram for NRZ under ESC Figure 3.28 Measured Q factor and Eye diagram for RZ under USC

101 The Q factor performance of PSK formats namely, DPSK, DQPSK and CSRZ- DQPSK modulations are also investigated at high data rates of 40 Gb/s, in a 16 channel DWDM system for a propagation distance of 600 kms in a NZDSF. Tables 3.14, 3.15 and 3.16 show the output Q factor measured for DPSK, DQPSK and CSRZ-DPSK modulation formats with balanced detection using the simulation parameters given in Tables 3.4 and 3.5. Table 3.14 Q factor measured for various input powers for ESC and USC Schemes under DPSK - 16 channel DWDM Power mw 0.25 0.5 0.75 1. 0 1.25 Q (ESC) 27 18 12 10 08 Q( USC) 32 20 14 13 11 Table 3.15 Q factor measured for various input powers for ESC and USC Schemes under DQPSK - 16 channel DWDM Power mw 0.25 0.5 0.75 1. 0 1.25 Q (ESC) 29 21 18 15 11 Q( USC) 36 22 17 16 13 Table 3.16 Q factor measured for various input powers for ESC and USC Schemes under CSRZ DQPSK - 16 channel DWDM Power mw 0.25 0.5 0.75 1.0 1.25 Q ( ESC) 29 22 18 15 11 Q ( USC) 37 24 18 17 13

102 It is found from Table 3.16 that the Q factor for the CSRZ-DQPSK modulation format shows better values at all power levels. Hence a combination of OOC based USC with the CSRZ-DQPSK shows better resistance towards the four wave mixing effect and gives better Q factor values. It is also inferred that at high powers the performance of DQPSK and CSRZ-DQPSK are similar and yield the same value of Q factors. Thus these two modulation formats perform equally well in the presence of non-linear effects. 3.6 IMPACT OF DISPERSION COMPENSATION ON MODULATION FORMATS Dispersion compensation is basically classified as: (a) pre-chirp techniques at the transmitter side, (b) dispersion compensation in the transmission line (in-line compensation) and (c) dispersion compensation at the receiver side, ( Keiser 2000 ). In pre-chirp, a chirp with the opposite sign of that of the fiber is introduced at the transmitter for reducing the GVD effects in the fiber.. The pre-chirp can be realized by several methods, namely by exploiting the internal chirp of the laser source (Wedding et al 1994) or of an external modulator (Gnauck et al 1991, Henmi et al 1994), by the implementation of complex transmitter structures using additional components such as phase modulators (Khosravani and Willner 2001). The impact of Dispersion slope for NRZ and DPSK modulation formats at higher bit rate has to be taken into consideration ( Castanon and Hoshida (2002). Hayee and Willner (1997) showed the concept of pre and post compensation of the fiber in the presence of non linearities and dispersion at the data rate of 10 Gb/s.

103 The main application area of these technique are the cost effective, optical short-reach systems (e.g. MANs) with smaller channel bit rates, but in combination with dispersion compensation techniques they can enable a performance improvement even in high-bit rate transmission systems over long distances (Sano etal 2000). In-line dispersion compensation is realized in the optical domain. This is achieved by chirped fiber gratings, using DCF fibers or phase conjugators. The post-chirp techniques at the receiver side are characterized by the compensation of the chromatic dispersion in electrical domain by Maximum Likelihood Detection (Otte and Rosenkranz 2000). In this section we have estimated the optimal length of dispersion compensating fibers to be used to realize a high Q factor for various input power levels. The simulation is done for 16 channels with the data rate of 40 Gb/s per channel propagating over a non-linear dispersive fiber, and the Q factor is determined for various modulation formats. The performance of the non-linear fiber is also studied at higher data rate for various modulation formats towards the improvement of Q factor. The impact of chromatic dispersion becomes larger with a system upgrade to higher channel bit rates greater than 10 Gb/s. The performance of NRZ, RZ and DQPSK modulation formats are analyzed with pre, post and symmetrical dispersion compensation techniques for a constant SMF length of 80 kms and a varying length of DCF. One loop in our simulation includes SMF of fixed length 80 km and EDFA 1 of gain 20 db, noise figure 4 db and the variable length DCF followed by EDFA 2 of 12.6 db gain. The other fiber parameters used are listed in Table 3.17. The receiver side of our simulation has a photo detector with low pass Bessel filter of cut off frequency 0.75 times the bit rate. Table 3.18 lists the Q factors measured for various lengths of DCF based post compensation for the NRZ, RZ and DQPSK modulation formats.

104 It is observed that the Q factor is maximized for a certain length of DCF and degrades if the length is increased or decreased from the optimum value. It is found from Table 3.18 that, the optimum DCF length for best Q- factor is different for different modulation formats. For example, in the NRZ format, a maximum Q factor of 13.66 is achieved for a DCF length of 15 kms. For RZ format the optimum DCF length is 13 kms giving a maximum Q factor of 15.79 and for DQPSK the optimum DCF length is 12 kms giving a maximum Q factor of 17.34. Table 3.17 Fiber parameters used in the simulation for studying the dispersion compensation schemes Parameters SMF DCF Length 80 km Variable Attenuation 0.2 db/km 0.25 db/km GVD 16 ps/nm/km -72 ps/nm/km Slope 0.08 ps/nm 2 0.08 ps/nm 2 Effective Area 80 µm 2 30 µm 2 Table 3.18 Q factor measured at Input power of 2 dbm - Post compensation using DCF SMF in km DCF length in km Q factor (NRZ) Q factor (RZ) Q factor (DQPSK) 80 10 12.29 14.59 15.11 80 11 12.98 14.88 15.23 80 12 13.08 15.67 17.34 80 13 12.45 15.79 15.45 80 14 13.27 15.26 16.12 80 15 13.66 14.79 16.33 80 18 12.33 14.23 15.67 80 20 12.00 13.37 15.76 80 22 11.10 13.19 14.35

105 Tables 3.19 and 3.20 list out the Q factors measured under Precompensation and Symmetric compensation by DCF. In the case of symmetrical compensation, the loop consists of a DCF of variable length, EDFA 1 of 12.8 db gain with 4 db noise figure, SMF of length 80 kms, EDFA 2 of 20 db gain, SMF of 80 kms, EDFA 3 of 20 db gain, a DCF of length 10 kms, and finally EDFA 4 of 12.8 db. Table 3.19 Q factor measured at Input power of 2 dbm - Pre- compensation using DCF SMF in km DC F length in km Q factor (NRZ) Q factor (RZ) Qfactor (DQPSK) 80 10 19.80 21.08 23.08 80 11 21.30 22.29 24.29 80 12 21.09 22.98 27.66 80 13 18.42 24.45 26.66 80 14 17.13 19.27 21.23 80 15 16.04 18.66 21.88 80 18 14.42 15.43 17.34 80 20 14.00 14. 60 16.35 80 22 12.10 14.10 16.11

106 Table 3.20 Q factor measured at Input power of 2 dbm - Symmetrical compensation using DCF SMF in km DCF length in km Q factor (NRZ) Q factor (RZ) Q factor (DQPSK) 80 10 15.75 18.08 20.08 80 11 18.98 19.29 22.29 80 12 16.34 21.98 24.98 80 13 14.72 16.45 17.45 80 14 13.98 16.27 18.27 80 15 15.69 13.66 15.66 80 18 13.89 12.33 14.33 80 20 12.99 11.00 12.00 80 22 11.08 10.10 13.10 It can be concluded from these investigations that, an optimum DCF length can give us the best Q factor performance and this optimum length is also dependent on the modulation format as well as the location of compensation. Comparing Pre, Post and Symmetrical dispersion compensation techniques in terms of Q factor realized, it is observed that Precompensation is the best option. It is also noted that the DQPSK modulation format shows the best Q factor performance. DQPSK with pre-compensation using DCF of length 12 km gives the highest Q-factor of 27.66. Figures 3.29, 3.30 and 3.31 show the eye diagram and Q factor for the NRZ, RZ and DQPSK modulation formats, under different dispersion compensation techniques.

107 Figure 3.29 Q factor of NRZ under post compensation using DCF Figure 3.30 Q factor of RZ under pre-compensation using DCF

108 Figure 3.31 Q factor of NRZ under symmetrical compensation using DCF 3.7 SUMMARY In this chapter a modulation format has been identified by investigating its performance under Equally Spaced Channel and Unequally Spaced Channel schemes for DWDM systems. Unequally Spaced Channel assignment is studied using optical orthogonal coding technique. In addition, the impact of filtering techniques, modulation formats and dispersion compensating techniques in combating the linear and non-linear fiber impairments are also analyzed in terms of Q factor and Eye Opening Penalty to identify methods for improving the spectral efficiency. The right choice of the optical filter used for transmitter and receiver side is crucial especially in a system requiring more spectral

109 efficiency. The Q factor measured at the receiver output for various pretransmission filters like Bragg Grating, Gaussian, Super Gaussian and Fabry Perot Filters shows a reduction in Q factor for increasing distances indicating the necessity for a dispersion management technique. Super Gaussian filter of higher order, is observed to give the highest Q factor with Bragg grating Filter showing a closer performance. In terms of Q factor and EOP measurements, it is concluded that RZ based formats are more non-linear tolerant. In addition it is also concluded that under constant input powers the Q factor values are higher for USC scheme than the ESC scheme, irrespective of the modulation format. This is a significant inference implying that USC schemes are more FWM tolerant. Further, the comparison of Q factor values suggests that a combination of orthogonal code based USC and CSRZ-DQPSK formatting can effectively combat FWM even at higher channel powers. The Q-factor comparison of PSK based formats namely DPSK, DQPSK and CSRZ-DQPSK suggests that DQPSK and CSRZ-DQPSK have similar performance and are better FWM tolerant. The performances of the dispersion compensation schemes like post, pre and symmetrical compensation techniques are observed to be dependent on the modulation format as well as the DCF length. It can be concluded from our investigations that, an optimum DCF length can give us the best Q factor performance and this optimum length is also dependent on the modulation format as well as the location of compensation. Comparing Pre, Post and Symmetrical dispersion compensation techniques in terms of Q factor realized, for different modulation formats it is concluded that DQPSK modulation format with pre-compensation using optimum DCF length is the best option.