“…Following the procedure described in the papers [5][6], the power of new optical waves of frequencies f ijk = f i + f j -f k generated through the FWM nonlinear interaction of three channels f i , f j and f k at the end of a dispersion compensated section can be derived as (1) where P i , P j and P k are the input light powers with frequencies f i , f j and f k , r is the fiber refractive index, l the wavelength, c the light velocity in vacuum, c the thirdorder nonlinear electric susceptibility (6 · 10 -14 m 3 W -1 ), d ijk is the degeneracy factor that takes a value of 3 (for i = j) or 6 (for i ≠ j), A e1 is the effective cross-sectional area of fiber with length L 1 and attenuation coefficient a 1 to be compensated, A e2 is the effective cross-sectional area of dispersion compensating fiber with length L 2 and attenuation coefficient a 2 , and h ijk the FWM factor defined by (2) where the phase mismatch of the first fiber Db (1) ijk and that of the second fiber Db (2) ijk can be expressed in terms of dispersion and frequency differences as follows where D (m) is the fiber dispersion, dD (m) /dl is the dispersion slope and f 0 (m) (m = 1, 2) is the optical frequency that corresponds to a zero-dispersion wavelength of a corresponding fiber segment. In a WDM system with N wavelengths channels with equal channel spacing, the total FWM power generated at the frequency f n at the end of M-th dispersion compensated section, in the worst case, assuming that the input power per channel is constant P i (0) = P 0 , l = 1, 2,…, N can be written as follows (4) where b = 1024p 6 c 2 /(r 4 l 2 c 2 A e1 A e2 ), and M = L/(L 1 + L 2 ) (with L being the total transmission distance) is the number of dispersion compensated links, that is the number of amplifiers.…”