What question did this study set out to answer?

The main goal is to create a deep learning framework for quickly extracting parameters of ESD diode models.

February 22, 2026Microelectronics Reliability2 citationsOpen Access

Deep learning-based TLP-IV and overshoot ASM-ESD parameter extraction for ESD diodes

FCFredo ChavezMacquarie University SKSourabh KhandelwalMacquarie University

Key Points

The main goal is to create a deep learning framework for quickly extracting parameters of ESD diode models.
Utilized deep learning algorithms for parameter extraction.
Generated over 30 million training data points using Monte Carlo simulation.
Preprocessed training data to train DL extractors for TLP I-V and overshoot parameters.
Evaluated DL extractors on measured ESD data.
DL extractors can produce ASM-ESD parameters in seconds.
Achieved good accuracy in modeling TLP I-V and overshoot characteristics.
Significantly reduced the time required for parameter extraction compared to traditional methods.

Abstract

This paper presents a fast and accurate deep learning (DL)-based methodology for extracting the parameters of the industry standard Advanced SPICE Model for electrostatic discharge diodes (ASM-ESD) that are critical for the analysis of electrostatic discharge (ESD) events. ASM-ESD is the industry-standard compact model for ESD diodes, and its conventional parameter extraction strategy requires expert knowledge, and it can take days to weeks to obtain an accurate set of parameters. The DL-based parameter extraction of ASM-ESD starts with generating over 30 million training data points via Monte Carlo simulation. After preprocessing, the data is used to train the transmission line pulse current–voltage (TLP I-V) and overshoot DL parameter extractors. The performance of the DL extractors is then evaluated on measured ESD TLP I-V and overshoot data. The results show that the DL extractors can instantly generate ASM-ESD parameters to model the TLP I-V and overshoot characteristics with good accuracy while greatly reducing the extraction time to seconds. This work shows the great potential of using DL for developing fast and accurate charge device models for circuit simulation to prevent machine-related failures due to electrostatic discharge. • DL-based framework for automatic parameter extraction of ASM-ESD Parameters for ESD diode circuit simulation. • Developed DL extractors simplifies the extraction process and can extract ASM-ESD parameters within seconds. • Extracted model captures the TLP I-V behavior and voltage overshoot over a wide range of conditions.

问 AI

Bookmark

View Full Paper