EFECTO DE LA ADICIÓN DE LÍPIDOS SOBRE LA PRODUCCIÓN DE METANO

In realistic long-haul transmission scenarios, the difference between the GN and EGN models in predicting system maximum reach is small, as discussed in Sect. 3.2. In order to achieve a substantial prediction difference between these two models, even in the presence of the measurement uncertainties typical of experimental set-ups, the link had to be specially designed. In our experiment, a very short span length (25 km) was chosen.

The experimental setup is shown in Fig. 6-1. An array of 19 lasers positioned between 192.916 THz and 192.268 THz was finely tuned at 36 GHz frequency separation. The CUT, at the center of the comb, was generated using an external cavity laser (ECL) while for all INTs distributed-feedback (DFB) lasers were used. A couple of single nested Mach-Zehnder modulators (SN_MZM) modulated the odd and even interfering carriers. Polarization multiplexing of INTs was obtained through a PM emulator, while the CUT was modulated by a double nested Mach-Zehnder (DN_MZM) that directly generated a polarization multiplexed optical signal.

We first considered a standard scenario where PM-QPSK was chosen both for CUT and INTs. The symbol rate was 32 GBaud and the electrical signals driving the modulators were generated using fast digital to analog converters (DAC). Four de- correlated 211-1 pseudo-random binary sequences (PRBS) were digitally filtered, in order to obtain a square root raised cosine spectral shape with roll-off 0.1 and electrical bandwidth equal to half the symbol rate. A digital pre-emphasis was also applied to partially compensate for in-band bandwidth limitations of the Tx components. The used DACs were CISCO prototypes running at 64 GSample/s (corresponding to 2 samples per

symbol) and characterized by four independent output ports; therefore, a single prototype was sufficient for generating the in-phase (IX, IY) and quadrature signals (QX, QY) for each polarization that drove the modulator of the CUT.

The INTs were modulated using only two driving signals since they were polarization- multiplexed using a PM-emulator. Therefore, a second DAC prototype was sufficient for the simultaneous generation of the in-phase and quadrature (IX1, QX1) signals for the even-channels and for the odd-channels (IX2, QX2).

Fig. 6-1: Experimental setup.

A second scenario was selected in order to highlight the change in maximum reach related to the statistical properties of the adjacent channels. This aspect is properly taken into account by the EGN model while it does not affect the GN model, since the latter is intrinsically modulation-format independent. Specifically, the INTs were generated with a Gaussian-distributed constellation of symbols. The launched optical power of the Gaussian INTs was adjusted to be the same as that of the PM-QPSK INTs. The signal samples were clipped at a value equal to 3 , in order to limit the penalty due to the

finite resolution of the DACs, and digitally filtered to achieve exactly the same Nyquist shaping and pre-emphasis applied to the binary PRBS sequences.

The WDM comb was launched into a re-circulating fiber loop that made use of EDFA-only amplification and consisted of four spans of uncompensated SMF, with length approximately equal to 25 km (see Fig. 6-1). A first variable optical attenuator (VOA Pch), inserted at the beginning of each span, was used to adjust the total launched power, while a second one (VOA Span) was used to force the total span loss to 18 dB. The average fiber losses were directly measured on the spools while dispersion and non-

linearity coefficient were taken from manufacturer data-sheets. Extra-losses due to splices between patch-cords and fibers at each side of the spools have been carefully characterized and taken into account in model predictions. All fiber characteristics are summarized in Table 6-1.

Table 6-1: Parameters of the fiber types Ls [km]  [dB/km]  [1/W/Km] D [ps/nm/km] Extra Loss Input [dB] Extra Loss Output [dB] Span 1 24.9 0.196 1.3 16.66 0.58 0.32 Span 2 25.7 0.192 1.3 16.66 0.60 0.37 Span 3 24.8 0.196 1.3 16.66 0.75 0.52 Span 4 25 0.189 1.3 16.66 0.56 0.59

The total launched power was controlled with a resolution of 0.2 dB exploiting the internal power monitor of the EDFAs and knowing the losses of VOAs and fibers. The loop also included a spectrally-resolved gain equalizer (GEQ) and a loop-synchronous polarization scrambler (PS) to compensate for the EDFA gain-tilt and to effectively average the impact of polarization effects, respectively. The GEQ was not able to correct gain tilt lower than 1 dB and gain ripples of any value. A fifth EDFA was used to compensate insertion losses due to GEQ, PS, coupler and acousto-optic modulators (AOM), that, as a whole, act as an extra artificial span with no dispersion.

At Rx, the WDM signal was first sent into a tunable optical filter with bandwidth 50 GHz and subsequently fed to a standard coherent Rx front-end, where the signal was mixed with the local oscillator, i.e., a tunable ECL different from the one used at the Tx. The four electrical outputs of the Rx front-end were digitized using a 100 GS/s real-time oscilloscope (composed of two synchronized Tektronix DPO73304DX). Offline DSP was used to down-sample, equalize and demodulate the acquired signals.

The span length was properly selected to only 25 km in order to emphasize the difference in max-reach prediction obtained by GN and EGN models, to be able to detect it reliably even in the presence of measurement uncertainties. At the same time we were forced to insert extra-loss in the spans in order to keep the maximum number of recirculations below 20 (i.e., 80 spans) and thus avoid excessive accumulation of residual gain tilt and other effects, such as PDL.