Comparison of BLER in Neural Receivers Utilizing a Novel OFDM Framework System

A framework system environment for neural receivers is described and developed for OFDM-based systems following standardized 3GPP scenarios. System performance is assessed using the Block Error Rate (BLER) metric. In addition, the effectiveness of neural receivers was evaluated against conventional approaches, including the baseline Perfect Channel State Information (CSI) and Least Square (LS) estimation, under the 3GPP Clustered Delay Line (CDL) channel model. Results demonstrate that, in an OFDM-based system, the convolutional neural network receiver exceeds the benchmark models in terms of BLER for QPSK, 16-QAM and 64-QAM modulations. Upon simplifying the neural receivers by removing a layer, an improved outcome is seen for QPSK relative to 16-QAM.