On the Fairness of the Performance Evaluation of Probabilistically Shaped QAM Signals

In this work, we compare the performance of probabilistically shaped (PS) QAM systems with reference uniform QAM under two conditions: (1) for the same information rate (IR) and the same FEC rate; (2) for the same entropy and the same FEC rate but resulting in different IRs. We demonstrate that unrealistically large system gains and improved tolerance to fiber Kerr nonlinearity of PS-QAM can be observed under second condition due to different IRs. We conduct a detailed analysis and prove that these gains are solely attributed to the reduced IRs. Also, we show that under the first condition of same IRs, more realistic shaping gains are achieved and slightly worse tolerance to fiber nonlinearity is observed. Hence, we recommend to use “the same IR” condition for comparison and the estimation of system gains by PS-MQAM.