Application of Neural Network Based Cascade-able Transceiver Model in Serial Link Simulation

This paper describes the work-flow of developing black-box transceiver models for channel cascading using feedforward neural network (FNN). Compared to transistor level models, the FNN model is independent of the circuit simulator, whereas the former suffers from slow Newton-Raphson iterations. Furthermore, the generation of FNN model requires no substantial knowledge in circuit design, making it more feasible to the end-users who work in another field.