In this section, the results o f the Q-factor measurements versus AA are described, extended to different dispersion maps, launch powers and system lengths to determine the impact o f these parameters on the absolute value o f Q and the slope AQ/AÀ. As before, 2 channels at lOGbit/s were transmitted over a single span o f 63km SSMF.
Dispersion: Both channels were modulated with PRBS and additional DCF was added to the link. Unlike with the dispersion-managed schemes discussed in chapter 4, the DCF fibre was placed before the EDFA and, hence, only compensates for the linear pulse distortion o f the link. Therefore, the residual dispersion following the generation o f XPM was not affected. As shown in Fig. 5.8, the Q-factor improved by AQ « +1.5 with respect to the uncompensated SSMF span, independent o f AÀ. In summary, the compensation for linear pulse distortion in the pre-compensated link increases the absolute values o f Q but does not change the variation o f XPM with A/l expressed by the slope AQ/AX.
o (0
10
9
87
6
0.0
0.5
1.0
1.5
2.0
2.5
3.0
M a x [ nm' ]Fig. 5.8 Q -factor vs. 1/A2 measured at lOGbit/s, 2 '^-l PRBS, ]3dBm /channel, receiver power: = -34dBm, (O ): SSMF fibre only, (■ ): D CF+SSM F, DCF before EDFA com pensated only for linear distortion
Chapter 5: Q-factor measurements 137
O ptical power: To confirm that the variation o f Q with AT is caused by XPM alone, the launched power into the SSMF fibre was decreased to 3dBm/channel and the resulting Q- factor was re-measured as a function o f AT. The results in Fig. 5.9 show very little variation from the constant value o f Q 7. This is consistent with the linear regime where the transmission is only limited by fibre dispersion and ASE noise.
Fig. 5.9
A X [nm]
Experim ental values o f Q vs. AÀ for 3dBm/channel launched into 63km SSM F fibre, PRBS m odulation at lOGbit/s
The Q-factor was calculated using the split-step algorithm for a range o f launch powers between 3dBm/channel and 13dBm/channel as shown in Fig. 5.10(a). As expected from the XPM -induced chirp, which increases with optical power, the variation o f the Q-factor increased for higher launch power and can be expressed by the factor \AQ/A(l/AÀ)\. The initial value o f the Q-factor for wide channel spacing was dependent on SPM and Gaussian noise. In the power-dependent calculations, the Q-factor was normalised to g(AT=1.6nm ) in all cases to highlight only the variation o f the slope AQ /A(\/AÀ) with launch power. For example, lOdBm/channel resulted in A Q ^ -0.3-A('l/ATj whilst this figure was increased to -f).l'A (\/A X ) for 13dBm/channel. This was due to the increased distortion o f the 'T -lev el o f the detected channel with launch power resulting in a decrease o f the eye-opening and Q- factor. For a given channel spacing, the penalty due to XPM was investigated as a function o f the launch power. At AT=0.4nm the impact o f XPM on the Q-factor was negligible. However, AQ= -0.5 was measured for lOdBm/channel launch power and was reduced to approximately AQ= -1.3 for 13dBm/channel. In Fig. 5.10(b) a linear decrease o f the Q-factor with launch power is shown, resulting in AQ/AP » -0.06/mW. This reduction is a result o f the linear increase o f the XPM-induced distortion ctxpm with launch power.
Chapter 5; Q-factor measurements 138 B ë O) 6) II O) < 0.0 0- 0 .2 5 - 0 .5 0 - 0 .7 5 - 1.0 0- 1. 25-
1.0
1.5
1/AÂ [nm '] 0.0-I
-0 .5 - II g O) 6 II 0 5 10 15 20launch pow er per channel [mW]
Fig. 5.10 (a) Calculated variation o f Q with Î/AÀ, 65km SSM F, two PRBS-m odulated channels at lOGbit/s, (■ ): 3dBm/channel (alm ost no X PM distortion), (O ): lOdBm/channel, ( ^ ) : 11.7dBm/channel, (t); 13dBm/channel
(b) Reduction o f Q-factor due to X PM as a function o f launch pow er at A/l=0.4nm
D istance: The length o f the SSMF span was varied between Z=20km and 80km maintaining a constant launch power o f ISdBm/channel, using lOGbit/s PRBS modulation in both channels. As a result, the XPM distortion increased with the span length due to increased PM-IM conversion in the fibre following the generation o f XPM. The calculated Q-factor decreased with increasing span length L, independently o f the channel spacing, at a rate tAQ/tsL « -0.046/km as shown in Fig. 5.11. Since the additional fibre only affects the linear PM-IM conversion, the total amount o f nonlinear phase distortion remains constant, as confirmed by the identical slope IaQ/!A(\/!AX) in all cases. However, IaQ/lA(\/ÈS.X) was only changed for T=20km where and for L<Leff, the negative slope was reduced since less XPM chirp occurred in the span. In the following section, the length-dependent Q-factor measurements are extended to multiple amplified spans where each span contributes to XPM distortion.
Chapter 5: Q-factor measurements 139 10 Z,=20km 9 8 L=80km 7 0. 5 1.0 1.5 2.0 2.5 O' 8 - 1/AÂ [nm ']
Fig. 5.11 Calculated variation o f Q-factor vs. AÀ due to PM-IM conversion, 2 channels, lOGbit/s PRBS, 13dBm/channel into SSMF link, length: Z-=20km, 40km, 65km and 80km, increased PM-IM conversion decreases Q -factor
In summary, wavelength-dependent Q-factor measurements at lOGbit/s over a single fibre span have shown that the Q-factor is reduced for narrow channel spacing due to increased XPM distortion. It was confirmed that Q decreases linearly with 1/AÀ. This is consistent with pump-probe measurements o f chapter 4 using the same fibre span which have shown that XPM distortion ct^pm is proportional to 1/AA
The negative factor AQ /A(\/AX), indicating the transmission system tolerance to XPM distortion, was investigated as a function o f systems parameters in both experiments and simulations. The absolute value o f this param eter increased with launch power, however, it remained constant when varying span length and linear dispersion compensation.
Chapter 5: Q-factor measurements 140