Receiver sensitivity depends strongly on modulation rate and format.
It is common engineering practice to describe OSNR with respect to broadband ASE noise in a fixed measurement bandwidth, typically 0.1 nm. In this case, the required signal to noise ratio should be roughly proportional to the bit rate.
If you reference broadband optical noise to the signal bit rate or receiver electrical bandwidth, then you are less dependent on bit rate, but still sensitive to modulation format (e.g. NRZ IMDD, RZ IMDD, DPSK, coherent BPSK).
i have a graph of osnr with a variable amplifier spacing.i just want to know the required osnr in receiver so i can recommend on the value of amplifier spacing.
if u have a table of osnr depending on modulation format, please give me.
I don't have hard values for the minimum OSNR of current 2.5 G products. Results will be sensitive to eye opening, transmitter extinction and signal powers (i.e. how strong the optical noise is compared with electrical noise in the receiver).
Leibrich et. al. publish simulated and measured BER vs OSNR for various modulation schemes at 40Gbit/s. They required 26 dB (0.1 nm) OSNR for 10-9 BER with NRZ on-off keying. At 2.5 Gb/s, 16 times slower, the equivalent OSNR would be 12 dB lower, i.e. 14 dB required OSNR. Leibrich's simulations did not include electrical noise in the receiver, accounting for some of the discrepancy between measurement and simulation. Their ratio of optical to electrical bandwidth is smaller than in your case, so this probably under-estimates the ASE-ASE beat noise penalty by 0.5 to 1 dB.
Prof. Z. Ghassemlooy quotes 13 dB OSNR for 10-13 BER in his Optical Amplifier lecture notes (slide 23), but does not state whether this is measured or theoretical.
A value between 12 dB and 16 dB seems reasonable, possibly somewhat higher for a directly modulated DFB laser and APD detector, which generally won't perform as well as a well-designed externally modulated laser with EDFA pre-amp and PIN-FET detector.