Reliable simulations of high-speed fiber optic links are necessary to understand, design, and deploy fiber networks.
Laboratory experiments cannot explore all possible component variations and fiber environments that are found in
today's deployed systems. Simulations typically depict relative penalties compared to a reference link. However,
absolute performance metrics are required to assess actual deployment configurations. Here we detail the efforts within
the Georgia Tech 100G Consortium towards achieving high absolute accuracy between simulation and experimental
performance with a goal of ±0.25 dB for back-to-back configuration, and ±0.5 dB for transmission over multiple spans
with different dispersion maps. We measure all possible component parameters including fiber length, loss, and
dispersion for use in simulation. We also validate experimental methods of performance evaluation including OSNR
assessment and DSP-based demodulation. We investigate a wide range of parameters including modulator chirp,
polarization state, polarization dependent loss, transmit spectrum, laser linewidth, and fiber nonlinearity. We evaluate 56
Gb/s (single-polarization) and 112 Gb/s (dual-polarization) DQPSK and coherent QPSK within a 50 GHz DWDM
environment with 10 Gb/s OOK adjacent channels for worst-case XPM effects. We demonstrate good simulation
accuracy within linear and some nonlinear regimes for a wide range of OSNR in both back-to-back configuration and up
to eight spans, over a range of launch powers. This allows us to explore a wide range of environments not available in
the lab, including different fiber types, ROADM passbands, and levels of crosstalk. Continued exploration is required to
validate robustness over various demodulation algorithms.