Analysis of Neural Spike Signal Transmission Characteristics in 3-D Heterogeneous Integration

With the continuous development of artificial intelligence, spiking neuromorphic chip is favored for its architectural advantages of memory computing integration. The neural spike signal is used as the transmission signal in the spiking neuromorphic chip, and its completely different characteristics from the digital pulse signal will inevitably lead to different signal integrity problems. In this paper, we model the common Through Silicon Via (TSV) and the interconnection structure model of redistribution layers (RDL) in 3-D heterogeneous integration, analyze the insertion loss and transient impedance of the two structures respectively, and finally perform time-domain transient simulations with neural spike signals of different rising edges in the two models to observe their transmission characteristics. The simulation results show that neural spike signals with shorter rising edges can generate more serious signal integrity problems such as loss, reflection, and delay in the RDL-TSV interconnection structure.