the maximum values for conventional 10 Gbit/s NRZ systems. Note also that PMD is here mitigated as fast as the polarisation demultiplexing operation (in the order of 1 µs), i.e., at least three orders of magnitude faster than the best optical PMD compensators. Naturally, further investigations based on a larger number of measurements would be required to accurately estimate the outage probability of coherent-based systems due to PMD and PDL.
Chapter 4
Impact of nonlinear effects on coherent detection
In order to upgrade optical fibre networks and to provide higher capacities, much attention is paid to investigate high bit rate transmissions with narrow bandwidth signals to increase the spectral efficiency of transmission systems.
Thanks to the development of GHz technologies for digital signal processing, 40 Gb/s QPSK associated with coherent detection represent a fascinating tech-nique since it offers better sensitivity to optical noise and enhanced tolerance to linear impairments. In the previous chapter, 40 Gb/s PDM-QPSK with coherent detection has been proposed to take advantage of the tolerance to linear impairments provided by coherent detection while increasing the spec-tral efficiency. Such a transmission system appears attractive to overlay on DWDM existing 10Gb/s infrastructures, but its tolerance to nonlinearities should be better investigated. At present, there is no knowledge of what are the most impacting nonlinear effects on such a receiver. As a first step towards WDM configuration, in this chapter is presented an experimental investigation on the impact of single channel nonlinear impairments on 40 Gb/s coherent PDM-QPSK. Moreover, as a coherent receiver allows for compensation of large accumulated amount of chromatic dispersion [50], we investigate the impact of dispersion management on nonlinear impairments. Finally, through optilux, a numerical simulator developed by Prof. Bononi research group, a numerical investigation of the impact of nonlinear phase noise is presented.
This chapter is organised as follows.
In section 4.1 experimental results shows the impact of nonlinear effects
Figure 4.1: Experimental set-up
on 10 GBaud QPSK and 10 GBaud PDM-QPSK. The impact of dispersion management is evaluated by comparing two systems: one with dispersion com-pensating fibre (DCF) and one with all-digital dispersion compensation.
In section 4.2 a numerical simulation setup is presented. A Monte-Carlo simulation is used to calculate the performance of a multi-span optical system with 10 GBaud and 20 GBaud QPSK modulation.
In section 4.3 numerical results are reported. Single channel and WDM with 7 channel cases are analysed in order to quantify the impact of nonlinear phase noise.
4.1 Experimental Setup
As depicted on Fig. 4.1, the test-bed consists of 2×10 100 GHz spaced lasers combined by one multiplexer. A specific wavelength allocation is used in order to have CW channels only on the edges of the C-Band that maintain consistent amplifier loading. As shown by the optical spectra of Fig. 4.1, in the centre of the C-Band, the studied channel at 1545.72 nm consists of a narrow linewidth laser modulated either with 40Gb/s PDM-QPSK or with 20Gb/s single polarisation QPSK. The modulated signal passes through a Variable Optical Attenuator (VOA) and is combined with the 20 CW lasers before being injected in a booster. The comb is then sent into a
4×100km-4.1. EXPERIMENTAL SETUP 105
Figure 4.2: Q-factor evolution versus power ratio for PDM-QPSK signal and QPSK signal.
long standard SMF spans system, along which each double stage amplifier can include DCF. By adjusting another VOA at the input of the preamplifier, the OSNR is intentionally degraded at the receiver in order to measure BER in the range of 10−3 to 10−6. The signal is selected by a tuneable optical filter and sent to the coherent receiver. The coherent receiver has the same settings used in the experiments of section 3.3.
4.1.1 Tolerance to non linear effect with in line DCF
Constant output power Erbium doped fibre amplifiers (EDFA) are used in the experimental set-up. The output power varies between 16 dBm to 18.7 dBm depending on the EDFA. The power of the studied channel can be varied around the nominal power, defined as the power of each CW loading chan-nel, by adjusting the VOA at the transmitter side. The BER performance of 40Gb/s coherent PDM-QPSK is measured for various channel powers and is compared with the one of 20Gb/s QPSK signal. In both cases a polarisa-tion diversity receiver is used and the original polarisapolarisa-tion state recovery is
performed by a 9-taps adaptive filtering using CMA. The OSNR at the re-ceiver side was chosen to be 11.5 dB/0.2nm for 40Gb/s PDM QPSK and 8.5 dB/0.2nm for 20Gb/s QPSK signal. These OSNR values corresponds approx-imately to a back-to-back Q-factor of 13 dB in both cases. Fig. 4.2 shows the evolution of the Q-factor obtained from the measured BER as a function of the channel power variation. Straight line with triangles and dashed line with diamonds represent the performance of 10 Gbaud PDM-QPSK and QPSK respectively. We observe on Fig. 4.2 that both performance curves decrease when the channel power variation reaches 0 dB (e.g. channel power at same level as the one of CW channels). Moreover, PDM-QPSK outperforms sin-gle polarisation QPSK, more particularly for channel power variations higher than 3 dB indicating that single polarisation QPSK is more impaired by non linearities. This behaviour could be partially attributed to the 3 dB higher signal power per polarisation of single polarisation format. Therefore, this result demonstrates the interest of coherent PDM-QPSK to fully benefit from the enhanced complexity of coherent receivers while doubling the bit rate and slightly increasing the tolerance to nonlinear effects.
Figure 4.3: Q-factor performance versus channel power variation for 40Gb/s coherent PDM-QPSK, with and without in-line CD compensation.
4.1. EXPERIMENTAL SETUP 107
4.1.2 Tolerance to non linear effects without in line DCF
In a second experiment, in line DCF have been removed and cumulated disper-sion is compensated within the coherent receiver by using a 21-tap FIR filters.
Fig. 4.3 presents the comparison of Q-factor performance, obtained with and without in-line dispersion compensation, as a function of the channel power variation. We observed on this figure that similarly to the performance of coherent PDM-QPSK versus channel power variation without in-line compen-sation is drastically reduced when input power becomes higher than 0 dB, similarly to the one obtained with in-line compensation. However, this fig-ure also shows that the difference in performance between systems with and without in-line dispersion increases and becomes clearly larger than 1 dB.
4.1.3 Variation of transmitter OSNR and impact on non linear tolerance.
In a final step, an attenuator is inserted at output of the PDM-QPSK modula-tor, followed by a single channel EDFA and a 400 GHz-wide optical filter which selects the channel and remove the out of band noise. The previous experi-mental comparison is repeated with variable OSNR at the transmitter side set to 30 dB/0.2, 20 dB/0.2 and 15 dB/0.2nm, to further assess the tolerance to non linear effects of coherent PDM QPSK with and without in-line dispersion compensation. Experimental results are depicted on Fig. 4.4. Fig. 4.4.a shows that in the presence of dispersion management, the performance of coher-ent PDM-QPSK tends to decrease with the increase of nonlinear phase noise (NLPN), with penalties up to 2 dB. On the contrary, Fig. 4.4.b shows that uncompensated transmission gives stable performance whatever the increase of NLPN. Although this can be interpreted as a strong point of uncompen-sated systems it should be noted that the system with DCF is the one with the best absolute performance. Moreover, in Fig. 4.5 there is a curve that rep-resents the difference in Q-factor when the nonlinear mitigation algorithm [61]
is turned on. This processing can improve the Q-factor of 1 dB maximum and it is clearly visible in the constellation represented on the graph that the characteristic “beam” shape due to nonlinear phase noise disappear.
Figure 4.4: Q-factor performance versus channel power variation for various OSNR of 40Gb/s coherent PDM-QPSK, a) with and b) without in-line CD compensation.