Experimental investigations on power spectral density estimation in heterogeneous dispersion unmanaged transmissions

In this paper, we present and verify by experiments a semi-analytical model to estimate the power spectral density of the noise-to-signal ratio in dispersion unmanaged transmissions over heterogeneous fiber types. The model combines an experimental calibration and an analytical formula. After a one-span profiling calibration, the overall system performance is assessed by joining the calibration results with a cumulative summing formula, targeting as a performance estimate the spectral density of the noise-to-signal ratio. After recalling the fundamental theoretical developments, we report experimental validations for four-span long dispersion unmanaged heterogeneous testbeds. According to the experimental results, the estimation error on the signal-to-noise ratio is always below 0.3 dB. Moreover, thanks to spectral knowledge, we show that we can account for some digital signal processing performed at the transceiver, which impacts system performance, such as the carrier-phase estimation. We demonstrate that without altering the experimental calibration, we can predict the performance adaptively to the carrier-phase estimation implemented at the receiver, capping the estimation error to < 0.5 dB.