SMACD 2023

A tutorial on simulating Python defined ML device models in Xyce, an open source SPICE circuit simulator

Authors:
Paul Kuberry, Andy Huang, Biliana Paskaleva & Shahed Reza

Sandia National Laboratories, Albuquerque, New Mexico, US

Xyce is an open source, highly accurate, and massively parallel SPICE circuit simulator developed at Sandia National Laboratories. The Xyce Python Model Interpreter (Xyce-PyMi), also developed at Sandia, enables the execution of devices with behavior defined in Python as part of a larger circuit simulation using Xyce. This allows the circuit designers to rapidly generate models from data and to evaluate circuit performances in a production circuit simulator, leading to a robust and higher confidence design.

In this tutorial, starting with an introduction to Xyce, the audience will be guided through installation of Xyce-PyMi, development of an electrical device defined in Python, and will be given resources for use in developing more complicated data-driven device models using Python packages such as TensorFlow.

Design Automation of Analog and Mixed-Signal Circuits Using Artificial Neural Networks

Authors:
Nuno Lourenço¹, Nuno Horta¹ & José M. de la Rosa²

¹Instituto de Telecomunicações (IT), PT, ²Institute of Microelectronics of Seville, IMSE-CNM (CSIC/University of Seville), ES

This tutorial shows how to use Artificial Neural Networks (ANNs) for the optimization and automated design of analog and mixed-signal ICs. A survey of main (conventional) design methods and EDA tools is given to show the pros and cons of the prior art, as a motivation towards using ANNs as optimization engines. A step-by-step methodology is given to explain the key aspects to consider in the presented approach, such as dataset preparation, ANN modeling and optimization, and their application to a given design problem. Several case studies are considered as demonstration vehicles and examples at different hierarchy levels of analog IC design. The first one is based on the use of ANNs and spice-like simulators to optimize amplifiers (AIDAsoft) and the second one consists of combining ANNs and behavioral simulation (SIMSIDES) for the high-level sizing of Sigma-Delta Modulators (ΣΔMs). The tutorial is addressed to a general audience interested in learning the fundamentals and practical considerations of using ANNs as optimization engines for the automated design of analog and mixed-signal ICs.

Organizer:
Sarah Azimi

sarah.azimi@polito.it

Politecnico di Torino, IT

The focus of this special session is on the advancements in reliability and fault tolerance techniques for emerging technologies, including FPGA, SoC, MPSoC, three-dimensional technologies (such as 2.5D/3D Integrated Circuits), nanoscale FinFET devices, innovative memories (such as PCRAM, STTMRAM, RRAM), introducing state-of-the-art investigations on the methods and techniques as well as modern implementation technologies enabling an effective and efficient development of highly reliable systems.

The topics of interest include but are not limited to all the following topics:

• Design for Reliability
• Design for Single Event Effect Hardening
• Single Event Effect Modeling, Analysis, and Mitigation
• Design and Test for 2.5D/3D ICs
• Reliability of embedded systems such as FPGAs, SoC, and MPSoC

Organizer:
Juan Nuñez

Instituto de Microeletrónica
de Sevilla, ES

Edge Computing applications require platforms to extract task-relevant information from increasingly large amounts of data with stringent constraints on energy efficiency, performance, cost, and reliability. Traditionally, silicon CMOS scaling has been relied upon to meet these energy and delay constraints. However, the energy and delay benefits achievable via scaling are diminishing. Therefore, there is an urge to explore new energy efficient devices and interconnects as the building blocks for energy efficient systems. This special session aims to bring together contributions focusing on different topics related to the modelling, design of circuits and systems using emerging technologies, as well as to highlight the latest advances in novel computing paradigms.

Organizer:
Vasileios Ntinas

vasileios.ntinas@tu-dresden.de

Technische Universität
Dresden, DE

Today's computer systems are on the verge of a paradigm shift from general-purpose architectures to cutting-edge specialized technologies optimized for particular computational workloads. This trend is driven by the increased need for autonomous, high-performance systems that can sense and process remotely on a tiny energy budget and footprint. Emerging memory technologies are increasingly growing since they provide novel nanoelectronic devices with outstanding features for single electronic components, including non-volatility, adaptability, low-power operation, and ultra-dense integration. Their operating principles arise from various physical mechanisms, based on their nanoelectronic structure, leading to a wide range of device types, ranging from ionic, and crystallographic (ReRAM, CBRAM, PCM) to magnetic, and ferroelectric devices (MRAM, FTJ, FeFET). Even though the original target of such technologies is memory, they demonstrate suitability for a wide range of novel computing and bio-inspired applications. However, the readiness of these technologies is yet far from large-scale commercialization and they are still facing fundamental issues in modeling, simulation, and circuit design. The main goal of this special session is to bring high-quality research on emerging technologies into SMACD ’23 and to create a forum for fruitful knowledge transfer between the traditional SMACD audience and researchers working on novel computing approaches that utilize emerging devices.

The topics of this special session focus on:

• Physical and numerical emerging memory device modeling
• Circuit design techniques for In-Memory Computing with emerging technologies
• CMOS integration of emerging technologies
• Variability- and reliability-aware design of emerging device-based circuits
• Emerging devices for bio-inspired circuits and systems

Organizers:
Juan M. Carrillo &
José L. Ausín

jmcarcal@unex.es

University of Extremadura, ES

CMOS technologies and circuit techniques have facilitated the development of innovative physiological sensing devices, ensuring the validity of medical information through analog and digital signal processing methods. Traditional trends in the use of these electronic devices for healthcare are being modified by new habits for healthy lifestyles, paying special attention to prevention and monitoring as key functionalities of the technological solutions. In the last years, a new trend towards extremely low-voltage and ultra-low power circuits is emerging, motivated in part by the interest in implantable or wearable devices for biomedical monitoring, sensors, the Internet of Things (IoT) networks, and other similar applications. This special session tries to gather novel contributions and challenges related to the use of low-voltage low-power integrated circuits in the field, especially for biomedical applications.

Topics should be related, but not limited, to: