ELECTRONICS LETTERS 9th November 2000 Vol. 36 No. 23 System optimisation of 80 Gbit/ s single channel transmission over 1000km of standard fibre V.K. Mezentsev, S.K. Turitsyn and N.J. Doran The authors have demonstrated by numerical optimisation of a symmetric dispersion map and using nonlinear chirped return to zero modulation format the feasibility of a record 80Gbit/s single channel transmission over 1000km of standard fibre. Introduction: Over the last decade, an individual channel data rate has continued to grow. Transoceanic transmission distances have already been achieved in single channel experiments for fibre lines based on dis- persion shifted fibres at a bit rate of 40Gbit/s [1]. In addition, the prob- lem of increasing a data rate and error free propagation distance in standard monomode fibres (SMF) is becoming very important because of the immediate application to the upgrade of existing terrestrial links. Recently, successful 40Gbit/s transmission over 1000km of SMF has been reported [2]. In this Letter we report the results of the numerical system optimisation of 80Gbit/s transmission over 1000km of SMF. Transmission simulations: As a sample system for our study we used a symmetric dispersion map [3, 4] (as shown in Fig. 1) similar to the experimental setup [2]. A role of prechirping in such a map is minimised since chirp free points are located exactly at the beginning and at the middle of the map [3, 4]. Therefore, input pulse chirp is virtually elimi- nated from the list of major optimisation parameters. As a matter of fact, precompensation and postcompensation are tuned for largest propaga- tion distance after global system optimisation. The first half of the dispersion map starts with the SMF transmission fibre followed by the DOF and EDFA amplifier. The second half is a reverse mirror of the first half. Fibre parameters are listed in Table 1. The two EDFA amplifiers are deployed symmetrically to provide a com- pensation of energy losses. The inline optical bandpass filter with band- width 1.4THz follows each amplifier. A system performance analysis was carried out in terms of the maxi- mum propagation distance corresponding to a bit error rate (BER) less than <10 –9 . BER was estimated by means of a Q-factor. The data stream was modelled by periodic bit patterns consisting of a pseudo-random sequence of 128 or 256 Gaussian pulses. Pattern propagation over transmission and compensation fibres was simulated by a nonlinear Schrödinger equation with the effects of third- order dispersion and Raman gain included. The action of EDFA amplifi- ers was taken into account by introducing ASE noise to a signal with a typical noise figure of 4.5. To reach the maximum possible transmission distance, model system and pulse parameters were optimised. It was found that an initial pulsewidth of ~3ps provides the best system performance, and therefore most optimisations were performed for this pulsewidth. Results: An instant detection at every numerical step was performed to locate the maximum performance points inside the map period [5]. Fig. 2 shows an evolution of Q-factors along the transmission line for the longest error free transmission achieved. Sharp peaks located near amplifiers marked with vertical gridlines correspond to the best system performance. It turns out that peaks of maximum Q-factor gradually drift away from amplifier locations and eventually take off the amplifier location at Q = 6, which is required for error free detection. Such a walk off of the maximum performance points cannot be eliminated by prechirping of the initial pulse or by moving a launch point along the periodic map. An optimum detection offset depends on distance and average dispersion of the map. Detector offset tolerance can be expressed as the width of the peak at Q = 6 and it decreases with dis- tance starting at a value of 2km at the first few sections down to zero. Therefore the last couple of peaks are unlikely to be resolved in experi- ment because of low offset tolerance and BER being barely lower than the required value. The optimal detector locations are shown in Fig. 3. The optimal detector offsets are plotted in terms of accumulated disper- sion Dz = D(z Q – z n ) against number of periods past n, where D is the dispersion of SMF, z Q is the location of the maximum Q-factor and z n is the amplifier location. These results suggest a simple recipe of post- compensation to match an accumulated dispersion of the post-compen- sating fibre to the value Dz and to provide the best system performance exactly at the end of the line. To study the dispersion tolerance, the map average dispersion was varied by means of changing the dispersion of the SMF fibre. The reason for such an approach is to minimise optimisation of the energy balance between fibre losses and amplifier gain. However, in the experiment, a similar purpose is usually achieved by varying the length of the standard fibre or by adjusting the operating wavelength. Note that from the sys- tem optimisation standpoint both approaches are equivalent. Once the fibre configuration and pulsewidth are chosen, the only two essential parameters remain in the problem. One parameter, the average or residual dispersion, characterises the map, while another parameter, peak power, is a signal property. Table 1: Fibre parameters Fibre SMF DCF Dispersion [ps/nm/km] 15.8–17.3 –84 Dispersion slope [ps/nm 2 /km] 0.064 –0.23 Effective area [μm 2 ] 80 25 Loss coefficient [dB/km] 0.22 0.65 Fig. 1 Dispersion map Fig. 2 Q-factor against propagation distance Fig. 3 Optimal detector locations Necessary accumulated dispersion of post-compensation fibre (PCF) against number of sections passed by signal Average dispersion —■— 0 —▲— –0.09 —▼— 0.08