An Accurate Neuro-Space Mapping Method for Heterojunction Bipolar Transistor Modeling

In this paper, an accurate modeling approach using Neuro-Space Mapping (Neuro-SM) for Heterojunction Bipolar Transistor (HBT) is proposed. The novelty of this paper is to propose a new Neuro-SM model that adds the output mapping neural network and derive the relationship between input and output of the new Neuro-SM model. The neural networks can map the current and voltage signals from the coarse model to the fine model, and the proposed Neuro-SM model can modify the behavior of the coarse model to match the fine model. It can automatically adjust the input signals of the proposed Neuro-SM model through neural networks to accurately match the characteristics of the fine model. HBT modeling example in DC simulation, Harmonic Balance (HB) simulation, combined DC and HB simulation show that the new Neuro-SM model proposed in this paper can accurately model and improve modeling efficiency.