Modular Neural Network Based Models of High-Speed Link Transceivers

In this article, we address the nonlinear behavioral modeling of transceivers using feed-forward neural networks (FNNs) such that each modular block functions independently in a high-speed link (HSL) simulation. proposed technique, transceiver models are represented form kernel matrices, which values determined through FNN training. By feeding with information on voltages and protocols, time-domain HSL analysis is transferred to simple matrix multiplications, allows significant simulation speedup while preserving good accuracy. Compared conventional standards, IBIS or IBIS-AMI models, generation requires minimal effort, thereby permitting wider access technique. Furthermore, demonstrate highly robust flexible terms feature expansion. With minor adjustments advanced settings as equalization differential signaling can be easily included trained models.