pattern effect is negligible compared to the XPM-induced eyeclosure
at signal “1” and therefore, the CW probe method is
an effective approach.
Another approximation in this analysis is the omission
of pump waveform distortion during transmission. This may
affect the details of the XPM crosstalk waveforms calculated
by (13). However, the maximum amplitude of CJ, h(t), which
indicates the worst case system penalty, will not be affected
as long as there is no significant change in the pump signal
optical bandwidth during transmission.
In general, the impact of XPM crosstalk on the system
performance depends on the bit rate of the pump channel,
XPM power transfer function of the system as well as the
baseband filter transfer function of the receiver.
A. Waveforms of XPM Crosstalk
In order to understand the impact of XPM on the system
performance, it is helpful to look at time-domain waveforms
involved in the XPM process. As an example, Fig. 6 trace (a)
shows the normalized waveform (optical power) of the pump
channel, which is a 10 Gb/s (a7 - 1) pseudorandom bit pattern,
band-limited by a 7.5 GHz raised-cosine filter. The probe was
launched as a CW wave and its amplitude was normalized to
“1.” Due to XPM, the probe channel is intensity modulated by
the pump and the waveforms created by the XPM process for
two different system configurations are shown in Fig. 6. Trace
(b) in Fig. 6 is obtained for a single span system with 130 km
NZDSF, while trace (c) shows the XPM crosstalk waveform
calculated for a three-span system with 130 km NZDSF +
115 km NZDSF + 75 km standard SMF. Looking at these
traces carefully, we can see that trace (b) clearly identifies a
simple high-pass characteristic, which agrees with the similar
waveform been recently measured and reported in [6] in a
single span fiber system. However, in multispan systems, XPM
transfer functions are more complicated. Trace (c) in Fig. 6
shows that the amplitude of the crosstalk associated with
periodic “0101” pattern in the pump waveform is suppressed.