ECOSISTEMA MICROBIANO RUMINAL

The power transfer experienced by the channels during transmission influences the signal power profile along the fibre, and therefore it changes the nonlinear interactions due to the Kerr effect act on the sigal. As an example, the signal power profiles along an optical fibre of two channels in the presence of ISRS and a channel experiencing pure loss is shown in Fig. 4.14. Here it is possible to see that the power transfer occurs in the first section of the fibre span, exactly the section where the majority of NLI is generated. This implies that in the presence of ISRS, NLI generated during transmission is also affected by the signal power. For modelling purposes, the previous assumptions of Eq. (4.3) are no longer valid, and the dependence on power of η becomes necessary. An initial attempt to include this effect in GN model was performed in [60]. Subsequently, several derivations for a GN model that includes the ISRS effect were published [157–159], however they all lacked experimental validation.

To investigate the impact of ISRS on a transmission system, two different channels were evaluated using SNR as a performance metric. Firstly, the transmission was performed using a single transmission band: only C-band for the 1530 nm channel and only L-band for the 1600 nm channel. This scenario represents a transmission system not impaired by ISRS due to the limited bandwidth used. Subsequently, the complete spectrum was transmitted and the performance of both channels was evaluated as a function of input power.

Figure 4.15 shows the results for both transmission scenarios for the channel placed at 1530 nm and the three different distances under study. It is possible to note that the optimum power was shifted towards the higher powers as the fibre span increases in length. As described before, this effect is due to the larger ASE noise generated to compensate for the loss of a longer fibre span. For the 100 km span it was not possible to

−6 −4 −2 0 2 4 6 8 10 12 5 7 9 11 13 15 17 19 100 km 160 km 200 km Launch power (dBm) SNR (dB)

w/o L-band, experiment w/ L-band, experiment

w/o L-band, model w/ L-band, model

Figure 4.15: SNR as a function of launch power at 1530 nm. Markers show experimental data and solid lines represent the theoretical model. The transmission of the entire spectrum (9 THz) is shown in blue and the transmission of either only C-band or L-band channels are shown in red colour.

clearly distinguish the optimum power due to the effect of the noise from the transceiver subsystem. With a back-to-back SNR of 19 dB, this was the highest achievable SNR after transmission, and, therefore, saturation of the received SNR was found when the linear and nonlinear noise are smaller than this value. For the spans with lengths of 160 and 200 km the optimum power was found to be 6 and 8 dBm, respectively. Comparing this to Fig. 4.12, the channel located at 1530 nm experienced a power loss due to ISRS of 0.2, 1.5 and 2.6 dB for the 100, 160 and 200 km spans respectively. The depletion of the channel lead to a decrease in the NLI; however, due to the loss of power the optical amplifier after the fibre span needed to compensate a greater loss, introducing additional ASE in the process. For the spans of 160 and 200 km, the detrimental effect of the additional ASE noise can be seen before reaching the optimum power, with a reduction on the received SNR. As the signal power was increased, the channel experienced a stronger power depletion because of ISRS, the effect of NLI was reduced. This can be seen for all the evaluated distances, with the channels obtaining a higher received SNR after optimum power due to the ISRS effect.

At the other end of the spectrum the opposite effect was found. The measured SNR as a function of launch power is plotted in Fig. 4.16. The optimum powers for the three fibre spans when no ISRS was present were found to be 0, 7 and 9 dBm per channel. The difference, as compared to the 1530 nm channel, was due to the higher noise figures of the L-band EDFAs. As seen before, the 1600 nm channel experiences amplification due to ISRS. This amplification leads to a smaller gain required from the EDFAs after the fibre and, therefore, to a reduction of the ASE noise level at the receiver. For both spans of 200 km, an improvement in the received SNR before reaching optimum power

−6 −4 −2 0 2 4 6 8 10 12 5 7 9 11 13 15 17 19 100 km 160 km 200 km Launch power (dBm) SNR (dB)

w/o C-band, experiment w/ C-band, experiment w/o C-band, model

w/ C-band, model

Figure 4.16: SNR as a function of launch power at 1600 nm. Markers show experimental data and solid lines represent the theoretical model. The transmission of the entire spectrum (9 THz) is shown in blue and the transmission of either only C-band or L-band channels are shown in red colour.

was found when the channel was amplified. The higher signal power also increased NLI noise, and for all three fibre spans a stronger degradation beyond the optimum power was observed. At 1600 nm, the overall performance of the channels was improved due to ISRS, with the reduction in the ASE level and the increase in NLI noise reducing the optimum signal power towards lower powers by approximately 1 dB. The received SNR was increased by 0.3 dB for the fibre with a length of 160 km, and by 0.9 dB for the 200 km long fibre.

As seen from the results ploted in Figs. 4.15 and 4.16, a system employing large optical bandwidths will be severly affected by nonlinear effects. ISRS and NLI are closely related, with ISRS influencing the generation of NLI. To design and plan the optical networks for the future, a precise channel model is required that includes all these nonlinear effects. In this work, the experimental results were compared to the modified GN model presented in [60]. The modified model uses the set of equations Eq (4.7) to calculate the power profile of a given channel i, and utilises an effective loss coefficient (αeff) that matches the actual effective nonlinear length of the channel under

test. The performance of the system is then given by a modified version of Eq. (1.1):

SNR_i= Pch P_ASE,i+ PNLI,i

, (4.8)

where PNLI,i= ηi(Pch)P_ch3, with ηi(Pch) described in [60]. The difference between

Eq. (4.8) and Eq. (1.1) is that the ASE and NLI noise terms depend on the location of the evaluated channel i on the optical spectrum. The main assumption of this model is that no noise is transfered from one channel to another due to ISRS interactions.

From the experimental results in Figs. 4.15 and 4.16 it can be seen that the used model correctly predicts the performance of the system, both in terms of ASE and NLI noise generation. The crosstalk-free ISRS assumption is, therefore, confirmed by the experimental data following the predictions from the model for the ASE noise level (PASE,i), with no additional penalties observed. This assumption would later be

independently verified in [160].