A Tunable Neural Network based Decision Feed-back Equalizer model for High-speed Link Simulation

This paper presents a model combining a feedforward neural network (FNN) with a recurrent neural network (RNN) to model Decision Feed-back Equalizer (DFE). By using the FNN as the mapping between the tap values and the dynamic behavior of the DFE, a complete model of the DFE can be constructed for channel simulation. The paper shows a 2-tap DFE example in which excellent agreement between the model generated by the proposed method and transient simulation can be observed.