CHAPTER 2 CHARACTERIZATION AND ANALYSIS OF OOK AND DPSK OCDMA SYSTEMS

Transcription

1 30 CHAPTER 2 CHARACTERIZATION AND ANALYSIS OF OOK AND DPSK OCDMA SYSTEMS Section 2.1 discusses the optimization of OSNR by properly selecting the power level and spectral characteristics of optical source for the given SSMF fiber attenuation and dispersion characteristics are discussed. Considering the fiber dispersions such as chromatic dispersion and polarization mode dispersion and nonlinearities of the fiber such as self phase modulation (SPM), cross phase modulation (XPM) and four wave mixing (FWM), the techniques to overcome these effects, EDFA fiber amplifiers and DCF compensation are included for the desired performance of the OCDMA system. Dispersion compensation techniques for the single mode fiber are suggested, which leads to the optimum channel and source conditions to maximize the OSNR. The section 2.2 and 2.3 present the issues for mitigation of MAI, noise, E/D using Super Structured Fiber Bragg Grating (SSFBG), detection and thresholding schemes for OOK OCDMA and DPSK OCDMA respectively. Optical and decision thresholding techniques and the corresponding receiver structures to optimize the BER performance with varying number of active users and the complexity involved are discussed. The DPSK OCDMA system with balanced detection and OOK OCDMA system with power detection are compared. The superior BER performance, higher receiver sensitivity, merits and demerits of DPSK OCDMA are presented.

2 OPTICAL SOURCE AND FIBER CHARACTERIZATION Dispersion and attenuation are the signal degradation effects observed in all optical fibers. Dispersion is compensated using dispersion compensating fiber (DCF) and attenuation is equalized by the erbium doped fiber amplifiers (EDFA). Various qualitative analyses are carried out in the simulation for the applications of OOK and DPSK OCDMA systems Optical Source and Parameter Selection Optical fiber communication systems often use semiconductor optical sources, light emitting diodes (LED) and semiconductor laser diodes (LD) because of the several inherent advantages offered by them. Some of the advantages are compact size, high efficiency, good reliability, right wavelength range, small emissive area compatible with fiber core dimensions and possibility of direct modulation at relatively high frequencies. The semiconductor materials with direct band gap energies such as In GaAsP, In P and GaAs are used for making optical sources. LED and LD are the suitable optical sources since they have adequate output power for wide range of applications and the output can be directly modulated resulting in high efficiency. Surface emitting LEDs operate as Lamberdian source with beam divergence of 120º in each direction. Edge emitting LEDs have a divergence of only about 30º. The spectral emission width of 40 nm of LED make it unsuitable for high bit rate system due to the high dispersion caused in the fiber. In spite of a relatively low output power and a low bandwidth of LEDs compared to lasers, LEDs are useful for low cost applications with data transmission at a bit rate of less than 10Mb/s over a few kilometers. Multiple quantum well (MQW) structures emit light at different wavelengths resulting in broader spectrum up to 500 nm and are useful for local area WDM networks. On the other hand laser diodes are coherent light source used to

3 32 couple sufficiently high power into multimode fiber (MMF) or single mode fiber (SMF). They are used for 100 Gb/s and more data rate applications. Narrow spectral emission width, high coupling efficiency and high output power level are the advantages of LD for high speed application. Bin Ni and James S. Lehnert (2005) analyzed the performance of an incoherent temporal spreading OCDMA system with broadband light sources. Dividing the impact of the non ideal light sources on the system performances, they had shown that the optimal range for the spreading code length when the available optical bandwidth and data rate are fixed. Thermal noise effects on the source were not analyzed. Georg Clarici (2007) presented the analytical model of laser diode for high speed applications. Simulation of single mode lasers reproduced all key characteristics for the analysis. Other types of laser diodes were not analyzed for compression. Martin Rochette et al (2005) evaluated the upper limit of the spectral efficiency of OCDMA systems with coherent sources. Spectral efficiency of 2.24x10-2 b/s/hz was achieved with a maximum BER of in the direct sequence and phase encoded OCDMA systems. The maximum spectral efficiency of OCDMA systems with coherent sources was at least a factor of 5 higher than the OCDMA systems with incoherent sources. However, the spectral efficiency of systems with incoherent sources decreased with increasing number of users. Broader spectral width reduces the spectral efficiency and at higher power levels the fiber nonlinearities degrade the system performance. To improve the system performance it is necessary to optimize the source characteristics.

4 33 The coherent light source is selected so as to support high data rates and to provide high spectral efficiency. LD operating in continuous wave is hereafter referred to as (CW) laser and is the suitable optical source for OCDMA. The CW laser with narrow spectral emission width of 1nm, line width of 0.1 MHz and emission power level of 0 dbm at nm is shown in Figure 2.1. Figure 2.1 Optical Emission Spectrum of CW Laser The emission spectrum is measured at resolution bandwidth of 0.01 nm and all polarization set to zero degrees. It is quite suitable for OCDMA applications and to overcome MAI. After phase modulation using LiNbO 3 - MZM the power level is reduced to dbm, which is taken as the transmitted power for the user. However in the simulation, the power level is adjusted using variable optical attenuators and EDFA; operating frequencies are varied in the range from 1540 to 1560 nm for multiple users.

5 Optical Fiber and Parameter Selection Optical fiber exhibits the least and uniform attenuation in the wide range of spectrum used for optical communication. The parameters of the standard single mode fiber (SSMF) used in all the simulation are as follows Attenuation of 0.2 db/km Wavelength range of 1400 to 1600 nm Dispersion of 17 ps/nm/km Dispersion slope of 0.08 ps/nm 2 /km, Differential group delay of 3 ps/km and Allowable nonlinear shift of 5 mrad SSMF offers a wide bandwidth and data rate in range of tens of THz. However the possible transmission data rate is reduced due to the limitations of electronics. Under these conditions, the pulse spreading due to dispersion mechanisms and fiber nonlinearities degrade the system performance in the upper band. Chromatic dispersion varies due to the exponentially decreasing nature of refractive index of the fiber core with increasing wavelength of operation. Higher wavelength components travel faster along the fiber leading to pulse spreading in the time domain. In the waveguide dispersion, the effective refractive index of the fiber varies and the effective refractive index is inversely proportional to the wavelength of operation. Because of the material and waveguide dispersion effects the total dispersion varies from 1 to 20ps/nm-km with zero dispersion around 1310 nm. The fiber nonlinear terms causes frequency mixing as four wave mixing (FWM), self phase modulation (SPM), cross phase modulation

6 35 (CPM). In the presence of large electric field, this becomes significant and generates a new wave introducing cross talk. Stimulated Brillouin Scattering (SBS) depends on the source power and line width. By proper selection of optical coherent source characteristics, fiber channel characteristics, data rate and launched power; the fiber dispersion and nonlinear effects are minimized within the tolerable level for the given fiber span. Distributed nonlinearities of optical fibers and OSNR limit the communication capacity. It was analyzed by Andreas D. Ellis et al (2010) and suggested techniques to improve the capacity. The techniques are (i) compensation of intra-channel nonlinearity either through link design or signal processing, (ii) optimization of OSNR through careful link design and phase sensitive amplifiers with spacing optimization. However capacity demands in access network cannot be met with WDM or dense WDM techniques. Ansgar Steinkamp and Edgar Voges, (2007) analyzed the influence of polarization dependent losses (PDL) on the statistics of PMD and statistical interdependencies between first and second order PMD. However, the effects of PDL are minimal for short haul OCDMA. All optical data format conversion to and from DPSK were proposed and numerically demonstrated by Jian Wang et al (2008). Multicasting, multi channel and ultrahigh speed (160 GB/s) format conversions were also demonstrated by simulation. However, the deviation to the actual application was not discussed.chromatic dispersion limited by the hybrid ASK-DPSK modulation format were studied by Jian Zhao et el (2007) for enhancing the transmission reach, also shown that electronic equalization of ASK and DPSK separately did not improve the CD tolerance of ASK- DPSK signal. Other modulation formats were not discussed. Comparative study of the Shannon channel capacity was presented for dispersion free,

7 36 constant dispersion and variable dispersion of fiber by Jau Tang (2006). Different approaches were only approximated to the input power levels. Hofmann et al (2008) worked on WDM PON for higher bandwidth of operation in the 1550 nm range with a vertical cavity surface emitting laser arrays. Channel bandwidth of 10 Gb/s over 20 km of SSMF was demonstrated. However, Laser arrays and maximum bandwidth of 80 Gb/s were the limiting factors that are not addressed in detail. Optical signal to noise ratio is to be optimized for the link losses and distortion using EDFA and DCF. The characteristics and the number of EDFA and DCF are selected based on the gain and dispersion compensation respectively for the desired link length of SSMF. Using dispersion shifted fiber (DSF) the dispersion can be made zero at 1550 nm with suitable negative or positive dispersion. The same effect is obtained with dispersion flattened fiber at 1550 nm but provides uniformly constant dispersion from 1300 to 1550 nm Compensation using fiber Bragg gratings for Single Pulse Dispersion not only broadens the signal pulse but also reduces the peak power as shown the Figure 2.2 (a-d). For the bit sequence , the bit rate of 40 Gb/s and launched power of 0 dbm into fiber are considered for the analysis. The Gaussian pulse with peak power of 0 dbm at 199 ps and width of the pulse being 55 ps at -60 dbm is launched into the fiber as shown in the Figure 2.2 (a). The peak power of -26 dbm centered at nm with the spectral width of 10 nm at -60 dbm of the spectrum is shown in Figure 2.2 (b). Figure 2.2 (c) shows the dispersion effects experimented with 10kms length of SSMF. The compensated signal using FBG is shown in Figure 2.2 (e) and 2.2(f). Launched spectrum itself has

8 37 the power level of signal 26 dbm as in Figure 2.2(b). The power loss of 2dBm shown in the spectrum is observed after 10 kms of transmission as in Figure 2(d) for the fiber attenuation of 0.2 db/km spectral loss was not compensated by the FBG as in Figure 2.2(f). Launched spectrum itself has the power level of signal 26dBm as in Figure 2.2(b). The power loss of 2 dbm shown in the spectrum is observed after 10kms of transmission as in Figure 2.2(d) for the fiber attenuation of 0.2dB/km.spectral loss was not compensated by the FBG as in Figure 2.2(f). (a) Waveform (b) Spectrum Figure 2.2 (a-b) Launched pulse in time and frequency

9 38 (c) Waveform (d) Spectrum Figure 2.2 (c-d) Dispersion effects for 10 km of fiber in time and frequency (e) Waveform (f) Spectrum Figure 2.2 (e-f) Dispersion compensation using FBG

10 39 FBG has many disadvantages compared to DCF in terms of complexity involved in making the gratings and stability of operation. Further the FBG introduces on an average 20dBm of noise in the compensated signal. DCF characteristics are incorporated during the fiber fabrication process with the required dispersion coefficient and attenuation for a given length of the fiber. In the following discussions of the sections of 2.1 the spectral characteristics not provided. Wavelength drift is not reflected in the spectrum due to dispersion. Due to the time shift introduced by FBG, the dispersion compensation left shifts the pulse. Bette et al (2008) demonstrated the wavelength dependencies of CD and differential group delay (DGD) to the fiber birefringence value and derived the gratings. However, did not provide the DGD modeling as it was complex Compensation using fiber Bragg Gratings for multiple pulses Scheme for enhancing the thermal sensitivity of the FBG was discussed by Budiman Dabarsyah et al (2007). Tuning of group velocity dispersion (GVD) dispersion slope and wavelength of operation were done by controlling the temperature distribution of the uniform fiber Bragg grating (FBG). Dispersion slope was tuned from to ps/nm 2 with the center wavelength at nm. However, the requirement of circulator and the changes of GVD as dispersion slope is tuned are the disadvantages of the scheme. Similar analysis but with the bit Sequence at the bit rate of 40 Gb/s and 0dBm of power level is carried out. Figures 2.3(a), 2.3(b) and 2.3(c) show the launched, distorted and recovered pulse sequence after compensation respectively.

11 40 Figure 2.3 (a) Launched pulse into the fiber Figure 2.3 (b) Distorted pulse after 10 km of fiber length

12 41 Figure 2.3 (c) Recovered pulse at the end of 10 km It is observed that the bits are not properly detectable. Worst situation occur for 100 km of fiber even with periodically compensating for dispersion using FBG. Therefore using dispersion compensation unit (DCU) dispersion compensation is effectively carried out Compensation using DCF for multiple pulses To compensate for the attenuation and dispersion so as to maintain the required OSNR and PCR at the receiver front end, an erbium doped fiber amplifier (EDFA) and dispersion compensating fiber (DCF) are used. Due to the performance limitations, single EDFA and DCF cannot serve the purpose. Therefore the following parameters of EDFA and DCF are selected for the link with two spans each of 50 kms. Each EDFA in Figure 2.4 provides a gain of 11.5 db and with a total gain of 23 db. The parameters of DCF are as follows

13 42 Length of the DCF is 5kms Attenuation is 0.3dB/km Dispersion Compensation is 850ps/nm/km DGD is 0.2 ps/km Each Span of 50 km consists of one DCF and one EDFA as shown in Figure 2.4 Two such DCF are used in this scheme and provides a total dispersion of 170 ps/nm/km. Figure 2.4 Symmetrical Dispersion Compensation for a span of 50 km Optical signal to noise ratio is to be optimized for the link losses and distortion using EDFA and DCF. The characteristics and the number of EDFA and DCF are selected based on the gain and dispersion compensation respectively for the desired link length of SSMF. Using dispersion shifted fiber, the dispersion can be made zero at 1550 nm with suitable negative or positive dispersion. The same are the effect with dispersion flattened fiber at 1550 nm but provides uniformly constant dispersion from 1300 to 1550 nm. For the bit sequence at the data rate of 10 Gb/s and with power level of 10dBm is carried out. The launched pulse into the fiber is shown in Figure 2.5, the broadened pulse is compensated using DCF and amplified using EDFA after10 km with overlapping for consecutive 1s is shown in Figure 2.6. The recovered pulse with compensation and amplification after the link span of 10 km is given in Figure 2.7.

14 43 Figure 2.5 The launched pulse into the fiber, clearly distinguishable at 0 dbm Figure 2.6 Broadened pulse with overlapping after10 km of fiber

15 44 The pulse is not distinguishable at 0 dbm as seen from Figure 2.7 and even at a higher power level (5 db) too it is barely distinguishable. Figure 2.7 Recovered pulse at the end of 10 km With successive 1s and at higher bit rate, the compensation using FBG fails to recover the launched pulse as shown in Figures 2.3 (b) and 2.3 (c). There is a difference in power level of 1 dbm. The peak to peak noise levels were reduced from -12 dbm to -6 dbm as shown in Figures 2.6 and 2.7. DCF and EDFA for each of the two spans are called as dispersion compensation unit (DCU). Using the DCU at the transmitting end of the fiber is pre compensation, at the end of the fiber is post compensation and at the mid of the total length of the link is symmetrical compensation. Simulation at different locations of DCU in the fiber is shown in Figure 2.8 for two power levels and data rates.

16 45 Figure 2.8 Comparison of eye diagrams for different data rates and power levels For each DCU and fiber spans of 50 km which forms a subsystem, the parameters and the responses are analyzed for the bit sequence with 10 Gb/s. The DCF parameters are: 5 km of length, 0.3dB/km of attenuation and dispersion compensation of 170 ps/nm/km with dispersion slope of 0.11 ps/nm 2 /km. DGD is taken as 0.2 ps/km. The parameters of EDFA in the DCU are: gain of 14 db, noise figure of 4 db, and wide noise bandwidth of 13 THz. The launched pulse after modulation is shown in Figure 2.9 (a). Compensated pulse after the first and second span by the DCU is shown in Figure 2.9 (b) and 2.9 (c) respectively.

17 46 Figure 2.9 (a) Gaussian pulse launched into the fiber with 8 dbm of power Figure 2.9 (b) Compensated pulses after the first Span

18 47 Figure 2.9 (c) Compensated pulse after the second Span The effect of total dispersion, power received and error performance of the system is obtained and the results are highlighted. The pulse with 10 dbm of power level at nm (0.01 nm Resolution bandwidth) is shown in Figure 2.10(a) for 10 Gb/s data rate.

19 48 Figure 2.10 (a) Spectrum of the Launched optical Pulse Each span of 50 km of fiber with dispersion compensation is carried out using DCU. The modulated spectrum and its power level is shown in Figure 2.10 (b), Figure 2.10 (c) shows the optical spectrum after MZM and 50 km of fiber. Figure 2.10 (d) and Figure 2.10 (e) shows the eye diagram along with Q factor variation and BER performance respectively. Maximum Q factor of , at the decision instant of bit period is achieved. The BER performance after DCU leads to improved power penalty.

20 49 Figure 2.10(b) Spectrum after MZM Figure 2.10(c) Optical Spectrum after MZM and 50 km of fiber

21 50 Figure 2.10 (d) The eye diagram and Q factor variation Figure 2.10 (e) BER Performance after DCU

22 51 Therefore for bit rate of more than 10 Gb/s and with the received power of less than 0 dbm, the symmetrical dispersion compensation leads to the best BER as shown in Figure Minimizing the Effects of Nonlinearities The fiber nonlinear terms cause frequency mixing as four wave mixing (FWM), self phase modulation (SPM) and cross phase modulation (CPM). In the presence of large electric field, this becomes significant and generates a new wave introducing cross talk. Stimulated Brillouin Scattering (SBS) depends on the source power and line width. By proper selection of optical coherent source characteristics and fiber channel characteristics, the nonlinear effects are minimized within the tolerable level for the given fiber span. Analytical expressions for phase and amplitude noises for phase modulated optical systems due to inter channel four wave mixing (IFWM) were derived by Alan Pak Tao Lau et al (2008). However, the analysis was not focused for multiple channel access network applications. Similar analysis was done by Alper Demir (2007). Rene-Jean Essiambre et al (2010) described the method to estimate the capacity limits of optical fiber networks. The sources of noise, Kerr nonlinearity and mitigation of impairments were described and compared the capacity limitations. However, the channel capacity for access networks was not specifically addressed. PMD monitoring for phase modulated signal using DGD generated interferometer filter was demonstrated by Yang et al (2008). 0 to 100 ps of DGD with 20 db of radio frequency power variation in 20 Gb/s NRZ DPSK

23 52 system was derived, which was insensitive to 0 to 640 ps/nm of CD. However, PMD monitoring technique in access network was not discussed. The fiber nonlinearities and dispersion related issues are studied to select the power level required for the system. As long as the optical power within an optical fiber is small, the fiber can be treated as linear medium. When the power level is high, the impacts of nonlinear effects are to be considered. Two parameters contributing to this are refractive index related parameters and scattering related impairments. Some nonlinear effects occur in multi channel WDM systems where interaction of signals at different wavelengths is possible. SPM and XPM affect the phase of signals and cause spectral broadening, which in turn leads to increases in dispersion penalties. SBS and SRS provide gains to some channels by depleting power from other channels. The nonlinear interaction depends on the transmission length and effective area of the fiber. SPM is a significant consideration in designing 10 Gb/s systems, and it restricts the maximum channel power to below a 10 dbm. XPM becomes an important consideration when the channel spacing is tens of GHz. FWM efficiency depends on signal power and dispersion, as well as channel separation. If the channel is close to the zero dispersion wavelength of the fiber, considerably high power can be transferred to FWM components. Using unequal channel spacing can also reduce effect of FWM. These findings are confirming with the results of earlier work done in analyzing the optical fiber characteristics. Dispersion plays a key role in reducing the effects of nonlinearities. However, dispersion itself can cause intersymbol interference. In the following example, the effects of dispersion compensation on system

24 53 performance in a high power regime where nonlinearities are active are considered. Two different versions are considered for analysis. In the first version, the system residual dispersion is 0, whereas it is 800 ps/nm in the second version. The transmission link contains 5 spans and the bit rate is 10 Gb/s. The dispersion at the end of 100 km SMF is 1700 ps/nm/km and its effective area is 72 square microns. The dispersion of DCF is -80 ps/nm/km. A 20 km DCF is used for the first version to totally compensate the dispersion. For the second version, an 18 km DCF is used to leave some residual dispersion after each span and this add 800 ps/nm total residual dispersion to the system. The effective area of DCF is 30 square microns. The loss in SMF and DCF are compensated by an EDFA with 25 db of gain for the first version and with 24.4 db gain for the second version. The Noise figure of the EDFA is 4 db. It is found that after about 10 dbm average power, SPM becomes a limiting effect. Figure 2.11 shows the eye diagrams for two different residual dispersion values and three different signal powers. when system residual dispersion is a) 0, b) 800 ps/nm. Experimented with 10 kms, 50 kms and 120 kms of fiber, hence it is stated in the respective places. In the entire situation, the DCF is at the middle (half way between the Tx and Rx) of the fiber link distance. The best performance are obtained with DCF in the Middle (Symmetrical).This is compared in the Figures 2.8 for 8 to -16 dbm of power levels and for 2.5 Gb/s and 10Gb/s.

25 54 Figure 2.11 Eye diagrams of the received signal for several received signal powers when system residual dispersion is a) 0, b) 800 ps/nm for one channel system This simulation shows that effect of SPM is reduced by incomplete compensation of the dispersion. It is observed that the power increase results in closure of the eye in the case of zero residual dispersion. For multi-channel system with 8 channels, the first channel is at THz ( nm) and the channels are separated by 100 GHz. SMF and DCF parameters are the same as in the previous example. To get more accurate results, nonlinear phase shift parameter of the fibers is set to a lower

26 55 value (3 mrad). Simulation results are shown in Figure The eye diagrams of the received signal for several signal powers with residual dispersion are 0 and 800 ps/nm respectively. Figure 2.12 Eye diagrams of the received signal for several signal power levels. a:0ps/nm of residual dispersion and b:800 ps/nm of residual dispersion for multi-channel system For an eight-channel system, the threshold power is approximately 10 dbm per channel. In this simulation, both SPM and XPM affect the system performance. The simulation also shows that nonlinear effects are reduced by local dispersion and better performance is obtained with nonzero residual dispersion.

27 OOK OCDMA SYSTEM The simplest technique of simulating the OOK OCDMA system consists in changing the signal power between two levels, one of which is set to zero and is often called on-off keying (OOK) to reflect the on-off nature is the resulting optical signal. Most digital lightwave systems employ OOK. OOK is identical with the modulation scheme commonly used for incoherent intensity modulation/direct detection (IM/DD) digital lightwave systems. For IM/DD systems, such unintentional phase changes are not seen by the detector, as the detector responds only to the optical power. The situation is entirely different in the case of coherent systems, where the detector response depends on the phase of the received signal. The implementation of OOK format for coherent systems requires the phase of the signal to remain nearly constant. This is achieved by operating the semiconductor laser continuously at a constant current and modulating its output by using an external modulator. Derivation of non linear equation for electronic dispersion compensation with OOK modulation using direct detection was carried out by Gilad Katz et al (2007). Suitability of the scheme for OCDMA system was derived. However, the same for DPSK and other modulations were not done. On off keying (OOK) modulation format is used in OCDMA system for payload data with power detection. The 10 Gb/s data rate through 100 kms of the fiber for OCDMA is very much affected by the attenuation and dispersion mechanisms of the fiber. The OOK OCDMA system is OSNR sensitive due to the power detection process at front end of the receiver. Hence the compensation for the fiber impairments is carried out to optimize system performance in terms of receiver sensitivity and BER.

28 57 OCDMA was the powerful alternative to TDMA and WDMA in fiber-to-the home (FTTH) Systems. Ken-ichi Kitayama et al (2006) demonstrated the OCDMA system architecture and its operation principle, code design, optical E/D using a long SSFBG. OCDMA over WDM PON was proposed. However, improvement towards MAI was not discussed. Wei-Ren Peng et al (2006) proposed frequency overlapping multi group scheme for a passive all optical fast frequency hopped (OFFH) OCDMA system based on FBG array with higher utilization of spectrum. Users were assigned the codes and divided into several groups with group interleaving. The interleaving of frequency allocations of different groups made the groups less correlated, and hence lowering the MAI. However, Gaussian profile of grating was chosen by Wei-Ren-Peng et al (2006) and did not suggest an optimum profile for power efficiency and MAI OOK OCDMA Schematic Description Optical laser diode operating at 0dBm power at 1550 nm is used to modulate the PRBS data using an intensity modulator followed by encoding using SSFBG. The encoded data from all the users are launched into the fiber through a star coupler. Other than the desired user, all the remaining users random data are to contribute for multiple access interference. The OOK-OCDMA signal gets attenuated at the rate of 0.2dB/km in the SSMF and also undergoes a dispersion of 17 ps/nm/km. To compensate for this, the parameters of DCF and EDFA are selected to ensure the required OSNR at the front end of the receiver. The threshold level is to set at slightly higher level due to the noise and interference of multiple users. When an optical threshold is used, the gain and dispersion of the threshold devices are adjusted to the required optimum value due to the fact that the optical threshold

29 58 removes the noise, MAI and beat noise to the required extent that the BER is maintained. The theoretical block diagram of OOK OCDMA with power detection system is shown in Figure 2.13 and the electrical domain receiver structure is shown in Figure Figure 2.13 OOK OCDMA system model Figure 2.14 Electrical Domain Receiver Structure Laser diode operating at nm corresponding to THz with output power of 1 dbm, line width of 0.1 MHz, without polarization reference and with zero phase generates the optical signal. The data is generated using an user defined code generator and precoded using one bit delay precoder to get the DPSK data and applied to the LiNb Mach-Zehnder modulator (MZM), and the modulated spectrum is shown in Figure 2.15 (b). The modulated signal is then encoded with a defined code in the encoder, for the signature of the user which is an 8 bit code of run length 2. This encoded output corresponds to the desired user. Other 7 such users operating at

30 to nm with channel spacing of nm (0.08 THz spacing). The 8 user encoded outputs are combined through a star coupler then to the fiber. The fiber lengths are varied from 50 kms to 600 kms to study the OSNR obtained at the input of the photodetector. The general parameters of the system are i) sample rate = THz, ii) number of samples = 4096, iii) samples per bit = 32, with a sequence length = 128 bits Encoded signal from the star coupler has the following characteristics and are shown in the Table.2.3. Single mode optical fiber (SMF) of length 100 kms with attenuation of 0.22 db/km, dispersion of 4.46 ps/nm/km and dispersion slope of 0.09 ps/nm 2 /km operating within the performance limit at 1200 nm to1700 nm is used in our experiments. Group velocity dispersion and third order dispersion are considered. Josep segarra et al (2007) proposed an all optical metro-access network using WDM/TDM architecture for PON based on optical burst switching. In the analysis it was shown that the design of optical network complicated the issues for the QoS.. Method for monitoring of simultaneous optical signal to noise ratio (OSNR) and chromatic dispersion (CD) in 40 Gb/s WDM systems were proposed and demonstrated by Lamia Baker- Meflah et al (2007). 20dB of dynamic range of OSNR was measured with 1dB of accuracy for the OSNR values of less than 20dB. The suitability of the scheme for access network was not discussed Multiple Access Interference Analysis Incoherent system of our proposed model uses 2-D coding schemes to provide better correlation performance and improved power and bandwidth efficiency in the asynchronous mode of operation. As analyzed by Xu Wang and Ken-ichi Kitayama (2004), the received optical field at the input of the

31 60 photodetector of the desired user with m interfering signal (0 < m K-1) is given in equation 2.1, where K is the total number of users in the system. E( t) m m 1 m = TCR d i i d i i= 1 i= 1 j= i+ 1 { P + P + 2 PP cos( φ ) + 2 PP cos( φ )} i j ij (2.1) = Data + MAI + PBN + SBN where P d and P i are data decoded and interfering power respectively. The values of the terms P i is not fixed, however a random variable fluctuating around its average P i leading to MAI. First term is the data signal, second term is MAI, third term is primary beat noise (PBN) and the fourth term is secondary beat noise (SBN); T C is the chip duration and R is the responsivity of the photo detector. Assuming the bandwidth of photo detector is larger than the frequency difference between the incoming signals and for smaller numbers of users the interference ξ = P / (2.2) i P d where,ξ is such that m ξ <<1, P i is the average interference power and hence the SBN can be neglected. For the incoherent system, the coherence time of light source (τ c ) is very much lesser than the chip duration (T C ) and cosine terms are equal to zero. The BN is a random process uniformly distributed over the interval [-π, π] during T C and therefore BN can be ignored by averaging throughout the detection. These conditions lead to approximating the system to be MAI dominant. The received signal is Z m { P d + P i.} T C = T C R + n ( t ) i 0 (2.3)

32 61 and n(t) is the photo detector noise included for the completeness of analysis. The average received signal Z is scaled by T C R and approximated as Z = (1 + mξ ) Pd ( T R) C (2.4) Assuming that the MAI and receiver noise both have Gaussian distributions, the error probabilities are derived as 1 P d ( D m ξ ) Pe (1 / 0 )( m ) = erfc (2.5) 2 2 σ 0 1 P d (1 + m ξ D ) Pe ( 0 / 1)( m ) = erfc (2.6) 2 2 σ 1 where 0 < D < (1+ mξ ) is the decision threshold; σ 0 and σ1 are the total noise variance with mark 0 or 1 respectively. σ = σ + σ + σ (2.7) MAI th s,1 σ = σ + σ + σ (2.8) MAI th s,0 where σ, σ and 2 MAI 2 th 2 σ s are the MAI, thermal, and shot noise variances, respectively. Therefore, setting the optimum threshold level is done (i) by selecting the large R L and nominal bandwidth to minimize the thermal noise and (ii) by selecting PIN photodiode for suitable responsivity and TIA to minimize the shot noise.

33 Optical Thresholder The decoding process at the receiver generates autocorrelation peaks from the designated user, and lesser db level cross correlation peaks from the non designated users. Optical thresholder (OT) is used to suppress the MAI and the arrangement consists of the following three components (i) Highly nonlinear fiber, (ii) Dispersion compensated EDFA (DC-EDFA) and (iii) Long pass filter to allow the desired wavelength range. Functional schematic of OT is shown in Figure Figure 2.15 Optical Thresholder DC-EDFA compensates for the dispersion and signal loss. One of the decoder as the desired user matching the code of the corresponding encoder in the presence of the selected number of interferers is applied to the optical thresholder.the nonlinear optical thresholder utilizes 500 m of highly nonlinear fiber (HNLF) with zero dispersion at 1553 nm. The dispersion of 0.19 ps/nm/km at 1550 nm, dispersion slope of ps/nm 2 /km at 1550 nm, effective area of 10 µm 2 and non linear coefficient of 20/(W.km) are the other parameters selected for HNLF. The HNLF shifts spectral power into longer and shorter wavelengths due to self phase modulation (SPM) and other fiber

34 63 nonlinearities. The longer wavelengths are passed to the receiver through the long pass filter. At the receiver, a power contrast ratio (PCR) of 25dB was measured between desired user and all the other interferers. The power measured at the input of the thresholder is the total received power in dbm. In the thresholder, the DC-EDFA average output power is set to 14 dbm for single user and the power penalty arises from pulse broadening, residual dispersion in the encoder and decoder. As the interfering users are added, the average output power of the thresholder is increased from the DC-EDFA to keep desired users average power into the HNLF constant. As the number of users increased, the thresholder output increases from 14dBm depending on number of users. The HLNF used here is polarization independent. The measurement emulates adaptive threshold detection consisting of a DC-EDFA, HNLF and a long pass filter, where DC-EDFA adjust its pump power level to optimize the threshold detection of the desired user signal while suppressing the interfering users signal. Each interfering user contributes a different amount of system penalty due to the differing MAI. Signal attenuation and dispersion leads to power loss and limits the transmission distance, OSNR and receiver sensitivity. Electronic amplification needs optical to Electrical and Electrical to optical conversion (O-E-O conversion). Electronic amplification depends on the bit rate and modulation format and hence electronic amplification is neither optically nor electrically transparent. All optical amplifiers are used as signal regenerators where loss is the limitation and single amplifier is be used for multiple channels and independent of modulation formats. Gain, bandwidth, gain flatness, noise

35 64 figure, maximum output power, coupling loss, pumping efficiency, polarization dependence and cross talk are the design parameters of an optical amplifier and are optimally selected. Gain spectrum of two level systems is Gaussian slope with peak at a particular wave length. Optical amplifiers impair the detection of phase modulated signals due to the interaction of signal and amplifier noise through the Kerr effect as described by Alan Pak Tao Lau and Joseph M. Kahn (2007). However, the interplay of chromatic dispersion (CD) and the Kerr effect on signal design and detection were not investigated. Polarization independent optical demultiplexing of 160 Gb/s optical TDM based data on cross-phase modulation (XPM) induced wavelength shifting in highly nonlinear fiber (HNLF). This has been experimentally demonstrated by Jie Li et al (2008). However, achieving polarization independency was difficult. Lijie Qiao et al (2007) described the scheme for maintaining constant output signal power in the presence of amplified spontaneous emission (ASE) noise for EDFA. The model was suitable for both single and multi channel operation for an input power range of 25 db and a gain of 0 to 37 db to operate the EDFA. Other than this range was not highlighted. Lee H. et al (2002) demonstrated an optical thresholder based on a short length of holey fiber to achieve enhanced code recognition quality in 255 chip 320 Gchip/s SSFBG based OCDMA code/decode system. The nonlinear thresholder was based on band pass filtering of spectrally broadened components generated by self-phase modulation (SPM) in an 8.7 m length of highly nonlinear holey fiber. However, it was not demonstrated for multi-user OCDMA system.

36 65 Performance of digital receivers with fixed and adaptive decision threshold were compared by Benjamin Puttnam et al (2008) in response to gain transients arising from network operation of gain-clamped EDFA. The advantages of adaptive decision threshold were given and the complexities involved were not analyzed. Anoma D. McCoy et al (2007) studied the semiconductor optical amplifier (SOA) based noise suppression for SAC-OCDMA system. The system had the limitation on optical filtering and not suitable for SAC- OCDMA applications. Thus, forcing the use of erbium doped fiber amplifier (EDFA) and noise suppression blocks. Waldimar Amaya et al (2008) presented a time spreading OCDMA system including non-perfect time gating and optical thresholding for OOK and DPSK modulation cases. Complexity involved was not analyzed. The carrier density fluctuations in semiconductor optical amplifier (SOA) imposed penalties on (i) PSK signals due to nonlinear phase noise and (ii) OOK signals due to self gain modulation. Francesco Vacondio (2010) proposed a scheme for equalization of impairments. Other amplifiers were not discussed. Satoshi Yoshima et al (2010) proposed a novel 10 Gb/s based PON over OCDMA system to realize full capacity of optical access network. The system was demonstrated by using multi level PSK, SSFBG, encoder/multi port decoder and burst mode receiver. However fiber dispersion and nonlinearities were not addressed in the analysis. As shown in Figure 2.16, the maximum of EDFA gain of 42 db occurs at 1530 nm with the flat gain region being 1540 nm to 1560 nm region. This feature of EDFA is useful for multiple channel amplification. The power conversion efficiency is 95.5% at 1480nm pumping and 63.2% at 980 nm

37 66 pumping for the signal at 1550 nm. Optimum length for doping is selected for amplification through population inversion as to the pumping signal gets absorbed along the length of the fiber. The dynamic range of operation is determined by the input signal power, output signal power and pump power for the required gain. Wavelength (µm) Figure 2.16 EDFA Gain variation with Wavelength of operation Figure 2.17 shows the gain variation with the input power. In the experiment, the EDFA gain of 14.5 db is always achieved for the input power of 1mW (0dBm). Also for 1mW of input power, 52 mw of maximum output power can be obtained as shown in Figure Figure 2.17 and 2.18 are plotted for 100mW of pump power. Gain up to 30 db is achieved with the pump power of 40 mw as seen in the Figure The amplified spontaneous emission (ASE) noise is directly proportional to gain and bandwidth and also the input signal. ASE is within 1dB in the operating wavelength and it is removed by properly selecting the gain and bandwidth and a filter is used to remove the noise outside the bandwidth. The ASE noise interferes with the detection process in the photo detector and contributes to the output introducing beat noise.

38 67 Figure 2.17 EDFA Gain variation with Input Power Figure 2.18 EDFA Output vs Input Power Figure 2.19 EDFA Gain variation with Pump Power

39 68 The Noise figure is given by SNR in / SNR out, and it is lesser with 1480 nm pumping than the 980nm pumping. Optical signal to noise ratio (OSNR) is maintained by properly choosing the design parameters for the given bit rates. The merits of semiconductor optical amplifier (SOA) are compactness, integration with optoelectronic components, functional applications and broad choice of operating wavelength. Input power was maintained constant at 400 mw for plotting the response. The pump power is varied from 1to 200 mw. The suitability of EDFA is seen by comparing the values with that of semiconductor optical amplifier given in the Table 2.1. Table 2.1 Comparison of Optical Amplifiers: EDFA and SOA Features EDFA SOA Typical Maximum internal gain 30 db-50 db 30 db Typical Maximum insertion loss 0.1dB-2dB 6 db-10 db Polarization sensitive Not sensitive to polarization Sensitive to polarization Pump source Optical Electrical Optical bandwidth 40 nm 30 nm Maximum output power 23 dbm 20 dbm Typical intrinsic noise 3-5 db 7-12 db OSNR Sensitivity In optimizing the simulation parameters, the fiber channel OSNR and Q factor decreases with increasing the distance as in the Figure As a consequence, the BER performance is improved with the OSNR. The rate of

40 69 change is compensated with suitably selecting the parameters of in-line EDFA and DCF; The OSNR of 36 db corresponding to the Q factor of more than 7 is maintained. (a) BER variation with Q factor (b) BER variation with OSNR Figure 2.20 Optimization of OSNR and Q factor for the Fiber

41 Disadvantages The electrical eye is unsymmetrical for both the cases. The dynamic decision threshold level setting in electrical domain is a complex problem as (i) The threshold setting has to change in accordance with the number of users (ii) Estimating the number of users in an asynchronous environment using digital signal processing or estimating the higher order harmonic levels of the received signal (iii) The optimum threshold set for an average number of users beyond certain level is not proportionate and the BER performance estimated with noise probability density function is not satisfactory (iv) Adaptively varying the threshold tends to increase enormously the cost, complexity in receiver preamplifier design and complexity in thresholder. And further burst errors degrade the system severely. (v) Alternatively using cost effective and lesser complex optical thresholding to mitigate MAI and noise gives more than 4 db OSNR improvements over the optimum threshold technique. Thus the performance of OOK OCDMA with optical thresholding is nearly the same as DPSK OCDMA without thresholding Simulation Schematic Simulation of OOK OCDMA and DPSK OCDMA is demonstrated using SSFBG E/D in Figure 2.21 shows the simulation setup for the demonstration and comparative investigation of DPSK OCDMA with OOK OCDMA. The schematic for generation of multiple access interference is shown in Figure The mode locked laser diode generated ~1.8-ps optical pulse at a repetition rate of 10 GHz with a central wavelength of nm.

42 71 Figure 2.21 Shows the arrangement for OOK OCDMA and with DCU The signal from the OOK modulator was split into two paths. The upper path is for target OCDMA user with 10 Gb/s OOK modulation using intensity modulator. The same schematic is used for DPSK OCDMA with phase modulator instead of intensity modulator. The lower one is used to generate different number and levels interferences. In the interference path, the MAI generator can generate interferences with different ξ 1 by tuning the variable optical attenuators (VOAs) and different K by adjusting the optical switches.

43 72 Figure 2.22 Simulation Schematic for MAI generation Here, ξ 1 is defined as the power contrasts ratio between single interference and the target signal, and K is the number of interferences. The signal and the interference are mixed and decoded by the decoder. At the receiver, a fiber based interferometer followed by a balance detector and a single PD was used for DPSK and OOK detection, respectively Super Structured FBG The encoders/decoders were written with significant frequency guard bands between frequency bins. These bins are partition of the available bandwidth into uniform segments for coding. The original intent of the guard bands was to reduce multiple-access interface (MAI) by assuring that when two codes do not have a given bin in common, no energy would leak through one bin to the other. Many codes are available, but long codes are usually considered. The assumption of square frequency response leads to the zeroing out of MAI after balanced detection when using constant cross

44 73 correlation codes. The encoders and decoders have identical codes, whereas the complementary decoder consists of the complementary code of the corresponding encoders. For a given bit rate, systems with greater optical bandwidth would offer better performance in terms of BER or, for fixed BER, could accommodate more uses for greater capacity. Since intensity noise is the principal noise source, and frequency guard bands reduce the occupied effective optical bandwidth and, therefore, the capacity. If ideal rectangular filters could be achieved, the encoders would carve out truly orthogonal frequency bins while exploiting all available bandwidth. Realistic FBGs that can be written in an effective manner will have finite roll off, leading to a tradeoff between MAI and intensity noise. This tradeoff requires the identification of further constraints on the code family to achieve optimization. However, simulation is achieved with FBG spectral responses with various levels of overlap between bins. These spectral responses are then used to predict the BER floor. The optimum spectral response is determined and a set of encoders/decoders is realized. In the experimental setup, encoding/decoding process is achieved by FBG that is working in transmission. The apodization profile A (z) of an FBG is the modulation index envelope that will be written in the fiber. The spectral response of a highly chirped FBG is simply an inverse translation of the gating apodization profile along the z-axis of the fiber. Basically, no modulation index (A=0) leads to a transmission bin, whereas a modulation index (A=1) leads to a non transmission bins, i.e., a reflective bin. An apodization profile based on super-gaussian lobes is used in order to minimize the ripples in spectral response that cause MAI.

45 74 The encoder is an SSFBG working in transmission that takes a broadband source and filters out all spectral content, expect those frequencies included in the users unique spectral code. All the users in the system share the same optical bandwidth and contain frequency elements for the same band; they access the channel asynchronous and without coordination. An N 1 coupler is used to combine all signals onto one fiber. The duration of the encoded signal and the decoded signal are about 800 ps and 1.6 ns, respectively. Therefore, in the interference path, the data rate is intentionally converted to 622 Mb/s to avoid the inter symbol interference. However as in the signal path the data are transmitted at 10 Gb/s data rate, the interference results in the performance degradation. But on the performance comparison, interference is considered as a fixed level interference to the received signal and neglected by taking all the measurement against the relative interference level ξ 1, which is proportional to the absolute interference level ξ Figures 2.23 to 2.35 show the simulation results for the E/D for the desired user for various conditions indicated in the relevant figures. Photodetector 1 (PD 1) is in the upper arm and photodetector 2 (PD 2) is in the lower arm of the differential detection scheme.

46 75 Figure 2.23 Encoded Spectrum after the 1 st SSFBG encoder Figure 2.24 Encoded spectrum at the input of Star coupler for the Desired user

47 76 Figure 2.25 Encoded signal at the input of star coupler Figure 2.26 Encoded signal at the input of Star coupler for the Desired user

48 77 Figure 2.27 Decoded signal after 1 st SSFBG Figure 2.28 Decoded Spectrum at the input of PD 1

49 78 Figure 2.29 Decoded Spectrum at the input of PD 2 Figure 2.30 Waveform at the input of PD2

50 79 Figure 2.31 Waveform at the input of PD1 Figure 2.32 Spectrum after Star coupler

51 80 Figure 2.33 Waveform after Star coupler Figure 2.34 Reflected spectrum from the 1 st SSFBG in the PD2

52 81 Figure 2.35 Reflected spectrum from the 2 nd SSFBG in the PD2 The schematic in Figure 2.34 shows the complete arrangement carried out for OOK OCDMA and programmable decoder for multiple users with DCU. This scheme simulates the OOK OCDMA system with intensity modulator and direct detection using PIN photodetector. At different stages of the proposed schematic, the signal and its spectrum are taken but their inferences are not given. Also, why the difference in the power level between signal and spectrum 2.3 DPSK OCDMA SYSTEM The use of PSK format requires that the phase of the optical carrier remain stable so that phase information can be extracted at the receiver without ambiguity. This requirement puts a stringent condition on the tolerable line widths of the transmitter laser and local oscillator. The line width requirement can be somewhat relaxed by using a variant of the PSK