Comparison between DWDM Transmission Systems over SMF and NZDSF with 25 4Gb/s signals and 5GHz Channel Spacing Ruben Luís, Daniel Fonseca, Adolfo V. T. Cartaxo Abstract The use of new types of fibre with high density and high capacity dense wavelength division multiplexing (DWDM) systems leads to the investigation of system performance with these fibres. This paper compares the performance of a high-density (.8bits/s/Hz) DWDM transmission system with 25 channels and 4Gb/s per channel with a channel spacing of 5GHz, using conventional single-mode fibre or non-zero dispersion shifted fibre (NZDSF). For this, semi-analytical simulation of the DWDM system is performed using rigorous models for optical multiplexer and external modulator. Nonlinear transmission phenomena in the fibre are also included in the simulation. The results show that NZDSF allows significant extension of reachable distances with dispersion compensation and in-line amplification. I. INTRODUCTION The increasing demand for capacity has led to the proposal of dense wavelength division multiplexing (DWDM) transmission systems and to the development of optical fibres that could support very high channel density and bit-rates. The generalized use of conventional single-mode fibre (SMF) with high dispersion of about 17ps/nm/km in present networks imposes serious limitations for bit-rates above 1Gb/s and channel spacing bellow 1GHz, mainly due to the presence of high dispersion (17ps/nm/km). The use of non-zero dispersion shifted fibre (NZDSF) with reduced dispersion of about 4ps/nm/km was proposed to reduce the limitations imposed by high dispersion. However, the occurrence of non-linear phenomena like cross-phase modulation (XPM) and four-wave mixing (FWM) may lead to additional degradation. In this work, a comparison between the use of NZDSF and SMF in a high-density (.8bits/s/Hz) DWDM transmission system. with 25 channels and 4Gb/s per channel with a channel spacing of 5GHz is carried out through semianalytical simulation. In order to take into account the main system impairments, rigorous models derived from experimental data for a modulator and an arrayed waveguide grating (AWG) demultiplexer are considered. R. Luís, D. Fonseca were with the Optical Communications Group, Instituto de Telecomunicações, Instituto Superior Técnico, Pólo I, 149-1 Lisboa. A. Cartaxo is with the Optical Communications Group, Dept. lectrical and Computers ngineering, Instituto de Telecomunicações, Instituto Superior Técnico, Pólo I, 149-1 Lisboa, Portugal. Ruben Luís, and Daniel Fonseca are now with Siemens Portugal, telephone: +351 21 4242561; email: ruben.luis@lis2.siemens.pt. This work was supported by FCT and POSI within project POSI/35576/CPS/2 - DWDM/ODC. II. A. Transmission System MODL DSCRIPTION The simulations performed are based in a point-to-point link. Fig. 1 presents a simplified diagram of the simulated system. In this model the optical transmitter generates 25 optical carriers spaced by 5 GHz, which are externally modulated by uncorrelated 4Gb/s sequences. The produced non-return to zero (NRZ) signals are multiplexed in an AWG and sent to the transmission path. In the receiver the signal is pre-amplified by a 15dB erbium doped fibre amplifier (DFA) and demultiplexed by an AWG before electrical conversion and processing. A channel spacing of 5GHz is used to achieve a spectral efficiency of.8bit/s/hz Table 1 presents the fibre parameters values used in this work for SMF, NZDSF and dispersion compensating fibre (DCF). LASR LASR LASR AWG NZDSF or SMF DFA Fig. 1 Diagram for the DWDM system Table 1 Fibre parameters AWG SMF NZDSF DCF Attenuation [db/km].225.25.5 Dispersion slope [ps/(nm 2 km)] 85 57-85 Dispersion [ps/(nm km)] 17 4-3 Non-linearity coefficient [W -1 km -1 ] 1.5 2 5.9 B. Transmitter The chirp and extinction ratio of the external modulator can affect significantly the overall performance of intensity modulated transmission systems. Due to this, a model for a modulator, which describes rigorously these effects in such a modulator, is considered. This model takes into account the amplitude and phase characteristics of the
modulator arms as well as the frequency limitations of the modulator. Fig. 2 - Representation of a modulator Fig. 2 presents a scheme of the modulator. This device was presented in [1] and is composed by an input Y junction connecting two arms with independent drive electrodes and an output Y junction. The input optical signal is splited in the input junction. The signal is affected in amplitude and phase as it propagates through the modulator arms, according to the applied voltage. The on or off states are obtained by constructive or destructive sum of the signals coming from both arms to the output junction. To obtain the characteristics of the modulator it was essential to evaluate the amplitude and phase characteristics of each arm. The electric field at the output of a modulator arm is given by [1]: () v = exp{ [ 1 2 α() v + j β()] v L} (1) where α and β are the power attenuation and phase constants. L is the interaction length of the arm, v the applied voltage and the input electric field. The non-linear characteristic of this device lies in the dependence of α and β on the applied voltage. This characteristic was determined in [1] experimentally. A set of these experimental points, shown in Fig. 3, was used to perform interpolation. 6. 5. 4. 3. 2. 1. attenuation [db] phase [rad]. -5-4 -3-2 -1 Voltage applied to arm [V] Fig. 3 - Attenuation and phase characteristics of one arm of the modulator With the dependence of α and β on the applied voltage it is possible to obtain the electric field at the output [1]: [ ( v, v ) = i SRin SRout exp{ [ 1 2 α( v ) + β( v )] L} + exp{ [ 1 2 α( v ) + j β ( v ) + φ ] L} ] j (2) v and v are the voltages applied to arms 1 and 2 respectively. i is related to the input electric field by [1]: i = ( 1+ SR ) ( 1+ SR ) in out (3) SR in and SR out are the input and output power splitting ratios respectively, φ is the phase difference resulting from the different lengths of the modulator arms. φ is radians for the conventional modulator and π radians for the π-shift modulator. The frequency limitations of the modulator are modelled by a 4 th order Butterworth filter with a 3dB bandwidth of 5GHz, since the experimental results described in [2] show an amplitude response similar to that one of that filter. In order to define the modulator model some assumptions are considered. The splitting ratios (SR in and SR out ) were considered identical, SR in =SR out =SR, in order to maximise the output power [1]. Additionally, in ref. [1] it is stated that the use of a π-shift (φ =π) modulator leads to better performance with SMF due to an exclusively negative chirp parameter. This can impose a significant time compression of transmitted pulses and therefore lead to system performance improvement. Fig. 4 shows an equivalent block diagram of the transmitter for each one of the DWDM channels. Data sequence v(t) lectrical Generator Voltage Converter v a (t) v (t) v (t) Butterworth Filter Modulator Optical Signal Fig. 4 - Block diagram of the optical transmitter. Slope and frequency limitations of the electrical generator are taken into account and lead to trapezoidal pulses with slope rate of 15% of the bit-time. The resulting signal is passed through the Butterworth filter (which describes the frequency limitations of the modulator) and follows to a voltage converter. This device produces the correct voltages of arms 1 and 2 of the modulator given, respectively, by: v v () t = VB 1 + vb 1 v() t () t = V + v v() t V and V are the bias voltages and v b1 and v b2 are the peakto-peak voltages of the modulators arms. From expression (2) the normalized optical power characteristic of the modulator can be obtained. Fig. 5 shows such a characteristic for a splitting ratio of 1.35. peak-to-peak voltage applied to arm 1 -.5-1 -1.5-2 -2.5-3 -3.5-4.6345.7542.4242.79545.9425.52955 b2.2592.567.83669-4.5-4.5-4 -3.5-3 -2.5-2 -1.5-1 -.5 peak-to-peak voltage applied to arm 2 Fig. 5 modulator output optical intensity as a function of the applied voltages to arm 1 and 2 (4)
C. Optical Receiver In the receiver, the WDM signal is amplified by an DFA before being demultiplexed by an AWG demultiplexer. The individual channels are then processed by the electrical part of the receiver. The influence of the amplified spontaneous emission (AS) noise generated by the optical preamplifiers on the system performance is assessed analytically as described in the following section. The model for the electrical receiver is based on the model presented in [3] for bipolar transistors with detection performed by a photodetector. This system includes the introduction of circuit noise. The circuit s frequency limitations were also considered to determine the construction parameters and, consequently, the noise characteristics. For further details revert to ref. [4-App.] D. rror Probability Model The error probability model is based on the Gaussian approximation with exhaustive formulation, i. e., the biterror-rate is determined by calculating the eye opening for the current at the input of the decision circuit for each transmitted symbol. The error-bit-rate is then given by: 1 F µ, l µ 1, l F P + e = Q Q (1) N al= σ, l al= 1 σ 1, l were µ i,l and σ i,l (i=,1) are the decision current mean and noise standard variations for the l th symbol at the optimum sampling instant conditioned to the bit i. F is the decision threshold, determined in order to minimize (1) and N is the number of symbols of the received signal. The noise components include signal quantum noise, AS noise; beat noise signal-to-as, AS-to-AS and thermal noise from the electrical receiver [4-Ap. A]. For the signal, sequences were used with 128 symbols[4-ap. B]. With multi-channel systems, the binary sequences in the even channels were inverted in order to uncorrelate the channels. III. LINAR TRANSMISSION The simulations performed with linear transmission were used to assess the degradation imposed by linear crosstalk and the combined effects of modulator chirp and dispersion. For this purpose a simple link without dispersion compensation and line amplification was investigated. In order to assess the performance degradation imposed by linear crosstalk, the bandwidth optimisation of the multiplexer and demultiplexer was performed. With the use of AWG filters the main concern is with signal distortion versus channel overlap. In [5] the optimisation of the AWG bandwidth was performed and a back-to-back link was considered for this effect. Assuming that the spectral width of the signal is not significantly changed by the presence of nonlinear transmission effects or chirp these results are expected to be hold in this study. A. Modulator optimisation The optimisation of the modulator was performed by extensive simulation. A fixed length of fibre was used to perform the simulations, in order to include the effects resulting from chirp. The polarization voltages V and V were chosen to allow a middle point in the normalized optical power characteristic of.5. This value would allow a maximum extinction ratio. The values chosen are V =-2.75V and V =-1.25V. The optimisation was done by changing the peak to peak voltages, v b1 and v b2. The modulator was optimised using a single channel system in a link with 17ps/nm, using SMF and NZDSF, respectively. For both fibres, considering linear transmission, the optimum peak-to-peak voltages were v b1 =- 1.4V and v b2 =.2V with a SR=1.35. Fig. 6 shows the evolution of the receiver sensitivity with total dispersion with and without using the optimised Mach- Zehnder modulator. Receiver -5-1 -15-2 -25 SM F with mod SM F without mod NZDSF with mod NZDSF without mod 25 5 75 1 125 15 Total Fig. 6 Receiver sensitivity for a single channel linear system as a function of total dispersion for NZDSF and SMF using the optimised modulator A degradation of 4dB in back-to-back with optimised modulator is present due to a limited extinction ratio of 9.3dB. The result of the optimisation was to value the impulse compression with chirp instead of back-to-back gain with an increase of extinction ratio. B. Simulation results -17-18 -19-2 -21-22 single channel multi channel NZDSF multi channel SMF -23 2 4 6 8 1 12 14 Dispersion [ps/nm] Fig. 7 Receiver sensitivity as a function of fibre dispersion Fig. 7 shows the evolution of the receiver sensitivity with total dispersion for a single and multi channel system, using SMF and NZDSF fibre and considering linear transmission.
The first remark is about 4dB of degradation in back-to-back due to linear crosstalk. This degradation is imposed by the high channel density (.8bit/s/Hz) and assumes already an optimisation of the AWG multiplexer and demultiplexer. In both situations (single and multi channel) the system performance begins by improving with the total dispersion due to the chirp introduced by the modulator. It can be noticed that for the single channel system the optimum point for total dispersion is 72ps/nm. There is no difference between the use of SMF or NZDSF because the performance is imposed by the total dispersion and not by fibre length. In the case of multi-channel systems the evolution of the system performance for SMF and NZDSF is almost identical up to 7ps/nm. From this point the degradation of the channels positioned in the far end becomes dominant due to dispersion slope and the difference of performance between using SMF or NZDSF becomes evident with a degradation of.4db in the latter case for the optimum dispersion point, around 12ps/nm for both fibres. IV. NON-LINAR TRANSMISSION The simulation of transmission with non-linear fibre is also necessary to evaluate the influence of non-linear phenomena in the system performance. In this section, the simulation results for a simple point-to-point link without dispersion compensation or in-line amplification are presented. A. Single Channel Transmission NZDSF -21.4-21.6-21.8-22. -22.2-22.4-22.6-22.8-23. -21.4-21.6-21.8-22. -22.2-22.4-22.6-22.8-23. SMF Fig. 8 - Receiver sensitivity as a function of fibre dispersion and transmitted power with SMF and NZDSF for single channel transmission Fig. 8 shows the evolution of the receiver sensitivity as a function of the total dispersion for single and multi channel systems using SMF and NZDSF, considering non-linear transmission. Output powers per channel from 7dBm to are used. For the single channel system using SMF the performance is nearly independent of the transmitted power for the range of optical power used. The results obtained are identical to those obtained for linear transmission. Using NZDSF a stronger dependence of the system performance on the transmitted power level is noticed which shows a stronger influence of self-phase modulation (SPM) in this fibre, as expected due to the higher non-linearity coefficient. In this case this influence is positive for the system performance, presenting a gain of the receiver sensitivity of 1dB at 15ps/nm for of transmitted power when comparing with SMF. At this point an increase on system performance of.2db is achieved with an increase on the transmitted power of 1dB. In the optimum dispersion point, 83ps/nm, the curves become almost coincident and the system performance can be considered independent of the transmitted power. B. Multi-channel Transmission Considering multi-channel transmission, this sub-section presents the simulation results for the same values of total dispersion and transmitted power per channel. NZDSF SMF -14. -15. -16. -17. -18. -19. -2. -21. -14. -15. -7dB m -16. -17. -18. -19. -2. -21. Fig. 9 Receiver sensitivity as a function of fibre dispersion and transmitted power with SMF and NZDSF for multi-channel transmission In the case of multi-channel systems additional non-linear phenomena have to be considered, evaluating the presence of FWM and XPM in the two types of considered fibres. In the case of SMF, due to a high dispersion, XPM is dominant. For NZDSF the presence of FWM is the main factor. Fig. 9 presents the evolution of the receiver sensitivity as a function of fibre dispersion and transmitted power with SMF and NZDSF for multi-channel transmission. The range of considered powers per channel is from 7dBm to. It is evident that the system performance for NZDSF is much more dependent of the transmitted power than for the SMF. This high dependence reflects the higher non-linearity coefficient present in NZDSF. However the optimum total dispersion is not power dependent in both cases, 123ps/nm for SMF and 11ps/nm with NZDSF. V. TRANSMISSION WITH DISPRSION COMPNSATION For the simulation of a transmission system with dispersion compensation and in-line amplifiers it was assumed that the optimum total dispersion of a link could be determined without considering SPM and XPM on the DCF. The optimum dispersion values determined in IV.B will be used to calculate the length of required DCF per span: Dopt nspan Ltr Dtr LDCF = (3) DDCF L DCF is the length of DCF per span and D DCF is its dispersion parameter. D opt and n span are the optimum dispersion for the link, and the number of in the link, respectively. L tr and D tr are the length per span and dispersion parameter of the transmission fibre.
A. Simulation results The transmission path is divided into sections. One span of transmission fibre, the corresponding compensation fibre and one or two in-line amplifiers compose each section. With one amplifier, the attenuation of the entire section is compensated. Otherwise, an optical amplifier to perform loss compensation succeeds each one of the fibres. As an example, to identify the structure of a section that begins with a span of SMF as transmission fibre, compensated by an optical amplifier and then a span of DCF followed by another amplifier the symbol SMF+AMP+DCF+AMP is used. Table 2 presents the results obtain for the simulations performed with dispersion compensation and in-line amplification. Links with 3km and 4km are considered with SMF and NZDSF. Table 2 Simulation results for transmission with dispersion compensation and in-line amplification. 3km 4km 3km 4km Structure of each Section Reachable no. of sections Receiver sensitivity [dbm] Power per channel [dbm] Length of DCF [km] SMF+DCF+AMP 3-15.76-4 5.516 DCF+SMF+AMP 3-16.42-4 5.517 SMF+AMP+DCF+AMP 3-15.8-4 5.517 DCF+AMP+SMF+AMP 3-16.65-4 5.517 SMF+DCF+AMP 2-16.71-4 7.276 DCF+SMF+AMP 1-2.24-4 6.553 SMF+AMP+DCF+AMP 2-15.45-4 7.276 DCF+AMP+SMF+AMP 2-16.8-4 7.276 NZDSF+DCF+AMP 7-11.81-5 1.196 DCF+NZDSF+AMP 6-15.3-5 1.152 NZDSF+AMP+DCF+AMP 7-1.47-5 1.196 DCF+AMP+NZDSF+AMP 6-15.7-5 1.152 NZDSF+DCF+AMP 3-16.75-5 1.451 DCF+NZDSF+AMP 3-16.35-5 1.451 NZDSF+AMP+DCF+AMP 3-15.42-5 1.451 DCF+AMP+NZDSF+AMP 3-16.1-5 1.451 In terms of noise, the configurations with two amplifiers have the advantage due to a higher optical signal-to-noise ration at the end of the link. Also when considering transmission fibre of 3 or 4km the latter case involves a higher noise power at the optical receiver due to the higher gain required on the line amplifiers. If considering non-linear phenomena the use of two amplifiers per section is not advantageous because the mean signal power along the fibre is higher. For NZDSF, were the system performance dependence with the transmitted power is stronger this factor can be dominant. Another factor to be considered is the signal distortion due to non-linear effects on the DCF. Because SMF has a high dispersion greater lengths of DCF are required, leading to higher distortion. The use of 4km SMF lead to a maximum distance of 8km. In this case the influence of AS noise is not significant because the number of in-line amplifiers in the link is low. The signal degradation due to non-linear effects prevents the signal from being recovered at the DCF. With NZDSF the number of reachable sections is 3, independent of the configuration used. However, the configurations with only one in-line amplifier per section have the best performance. This fact is due to the stronger dependence on the mean signal power along the fibre. With 3km the gain required for the in-line amplifiers is smaller, leading to less noise and smaller mean signal power along the fibre. In the case of NZDSF this means that a maximum of 7 sections can be reached. In this limit the influence of noise and non-linear distortion have equal weights in the overall system performance. The low dispersion of the transmission fibre allows a good signal recovery at the end of each section. For SMF the high dispersion combined with non-linear effects limits the extension of a link to 3 sections. VI. CONCLUSIONS The best results for SMF were obtained with 3km in a DCF+AMP.SMF+AMP with a receiver sensitivity of 16.6dBm. In this case a distance of 9km was reached, with 3. For NZDSF a link with 21km could be simulated, with 7 3km in a NZDSF+DCF+AMP configuration. The receiver sensitivity for this case was 11.8dBm. Considering the overall performance, the NZDSF fibre brings great advantages for DWDM communication systems in terms of reachable distance. Also the use of DCF becomes simplified considering that the lengths required are much smaller thus easier to integrate in a DWDM transmission system. VII. RFRNCS [1] Chris Lawetz, John C. Cartledge, C. Rolland, J. Yu, Modulation characteristics of semiconductor Mach- Zehnder optical modulator, Journal of Ligthwave Technology, pp. 697, April 1997 [2] R. Spickermann, S. R. Sakamoto, M. G. Peters, N. Dagli, GaAs/AlGaAs travelling wave electro-optic modulator with an electrical bandwidth > 4 GHz, lectronics Letters, vol. 32, no 12, pp. 1, June 1996 [3] G. Keiser, Optical Fibre Communications, McGraw-Hill, 1991 [4] D. Fonseca, R. Luís, Study of the realizability of multichannel optical transmission systems with external modulation of capacity >=1Terabits/s, Instituto Superior Técnico, Final Course Project, 1999 [5] D. Fonseca, R. Luís, Adolfo V. T. Cartaxo, Design and performance of AWG multiplexer/demultiplexer in DWDM systems, 3 rd Conference on Telecommunications, Figueira da Foz, Portugal, 21