Deep Simple Recurrent Unit-Based Transient Modeling Method for High-Speed Circuits

The need for rapid and precise transient simulation for signal integrity (SI) assessment of high-speed circuits in microwave systems becomes increasingly crucial. Conventional recurrent methods, including recurrent neural networks (RNNs), long short-term memory (LSTM) networks, and gated recurrent units (GRUs) have been employed for transient modeling. However, despite their accuracy, these methods face limitations such as high computational costs and constrained parallel computing capabilities, especially in high-speed circuits characterized by extremely long input bit patterns. To overcome these challenges, this article presents the Deep Simple Recurrent Unit (DSRU), which is a transient modeling method designed to predict the time-domain signal response of high-speed circuits. The DSRU method strategically segments input signals through windowing and processes them through multiple SRU layers, thus incorporating an improved intra-unit gating mechanism to enhance parallelism. The DSRU method is validated in two high-speed circuit examples, which demonstrate its accuracy and efficiency in the modeling and design of high-speed circuits compared to conventional RNN, LSTM, and GRU methods.