Thank u Mohamed sir, i think filter bandwidth should be equal to the bandwidth of your message signal. the idea suggested by you is a kind of hit and trial method and if filter the bandwidth is less than the bandwidth of modulated signal then data loss will be there and if it is more then noise content will be there. so is there is any mathematics related to bandwidth of modulated signal in this. what i think is, for example we have NRZ formatted signal at the data rate of R Gbps so the message bandwidth will be R/2. when it is modulated by DD-MZI, it is a kind of amplitude modulation where carrier signal is optical signal so the bandwidth become 2(R/2)=R. but as i use this in simulations, the best results observed at the bandwidth of 3R/2. And i am not getting the point to mathematically calculate it.