Wireless sensor networks offer flexibility to monitor environmental conditions. However, in some hostile environments with changing propagation occurrences, it is challenging to guarantee a network free from systematic errors (biases), those errors can corrupt the data gathered from the network. The authors intend to address this problem using a Bayesian approach. This study presents a new Bayesian wireless network model based on a belief propagation algorithm over a noisy communication, to seek the calibration of the network for correlated sources. They set up a point-to-point communication model, consisting of a frequency modulation signal at the transmitter and a discrete-time phase-locked loop structure at the receiver over an additive white Gaussian noise channel. Then, they use a Bayesian approach to model the propagation of information, and analyse the noise impact on their system. The simulation results show significant improvement within a few time instants in the mean squared error of the sensors internal states. An evaluation in a noisy environment confirms the robustness of the proposed system.