Nowadays, the enabling of 5G technologies has a significant interest. At the same time, fiber optical access networks, which are widely realized as passive optical networks (PONs, are considered to be the most appropriate candidates for enabling the 5G X-haul (x-front or back) . Thus, the demands on the integration of radio-over-fiber (RoF) technology in PON are growing significantly. The essential part of the implementation of each new technology (including converged RoF-PON) is providing the technical requirements for this technology. That is crucial for standardization, network planning and choice of the equipment. The advantages of mathematical modeling that can ensure separate investigation of different performance affecting factors can be effectively used for this purpose. Addressing this, we focus on the improvement of the modeling of the converged RoF-PON transmission system. In the current research, we analyze the modeling of the receiver sensitivity, which is one of the significant performance-affecting factors in PONs. The main goal of the research is to investigate, does the symbol-error-rate (SER) depend on the way, how the same receiver sensitivity is ensured. Respectively, to evaluate the impact of utilizing different values of the noise and bandwidth. Such investigation is crucial for the development of the models aimed to simplify the definition of the requirements to transmission systems characteristics related to noise (e.g. sensitivity, signal-to-noise-ration (SNR)). The investigation is performed for a 3.5 GHz quadrature-amplitude-modulation orthogonal frequency division multiplexing (QAM-OFDM) RoF signal. The analysis for the different number of OFDM carriers is provided. The error vector magnitude (EVM) which is widely used for the quality assessment of QAM signals is also utilized for quality evaluation.