Page 339 - Fiber Optic Communications Fund
P. 339

320                                                               Fiber Optic Communications


            As in Section 7.2, we assume that the optical filter is an ideal band-pass filter with bandwidth B = f , and
                                                                                              0
                                                                                          0
            the electrical filter is an ideal low-pass filter with bandwidth f . The variances of bit ‘0’ and ‘1’ are given by
                                                             e
            Eqs. (7.17) and (7.20), respectively as
                                                4k Tf      (    )
                                     2            B  e    2  eq  2
                                     = 2qI f +      + 2R      (2f − f )f ,               (7.131)
                                     0     0 e    R          ASE    o   e e
                                                   L
                                           4k Tf e   2 eq          eq
                                             B
                                2
                                = 2qI f +      + 2R   [2P f +   (2f − f )f ].          (7.132)
                                                                        o
                                                                            e e
                                                             in e
                                      1 e
                                1            R L       ASE         ASE
            The long-haul fiber-optic systems are typically amplifier noise-limited and, hence, some approximations
            can be made while calculating the Q-factor. The variance of shot noise and thermal noise can be ignored
            compared with the variance of signal–spontaneous beat noise. Also, when the signal power is large,
            spontaneous–spontaneous beat noise can be ignored as well. Under these conditions,
                                                 2
                                            2
                                            ≅ 4R P [Nn hf(G − 1)]f ,                       (7.133)
                                            1       in  sp        e
                                            2
                                            ≅ 0,                                           (7.134)
                                            0
                                               √
                                                       P in
                                            Q ≅                 .                            (7.135)
                                                  4Nn hf(G − 1)f e
                                                     sp
            From Eq. (7.135), we see that the Q-factor is independent of the responsivity R, under these approximations.
            Q can be increased by increasing P or decreasing the gain G. Since G = 1∕H, Q can be increased by using
                                        in
            low-loss fibers. From Eq. (7.135), we see that as the number of amplifiers (or n or G) increases, P has to
                                                                                            in
                                                                            sp
            be increased to keep the Q-factor at a fixed value. The enhancement of the signal power to counter the noise
            increase is known as the power penalty. Suppose the number of amplifiers increases from N to 2N, then the
            launched power should be doubled to keep the Q-factor fixed (or equivalently, BER fixed). In this case the
            power penalty is 3 dB. Using Eqs. (7.110) and (7.135), we find
                                                           2
                                                         Q f e
                                                 OSNR =      ,                               (7.136)
                                                          B opt
            wherewehaveused P = P ∕2 for OOK. Solid and dotted lines in Fig. 7.16 show the BER obtained using
                              in   in
            the exact Q-factor (Eqs. (7.131) and (7.132)) and approximate Q-factor (Eq. (7.135)), respectively. Since
                                       10  –5
                                      10  –10              P in = –6 dBm
                                     BER  10 –15

                                                           P in = –3 dBm
                                      10 –20


                                      10  –25
                                         50     60    70     80    90    100
                                                     No. of amplifiers

            Figure 7.16  BER vs. number of amplifiers for direct detection system. Parameters: n = 2,  = 0.2 dB/km, amp. spacing
                                                                         sp
            = 80 km, gain G = 16 dB, R = 1000 Ω, T = 200 K, R = 1A/W.
                                 L
   334   335   336   337   338   339   340   341   342   343   344