The Controller Design and Integration of Memristor-Based Neuromorphic System

Abstract

In this project, we designed and fabricated a neuromorphic system controller in TSMC 65 nm technology for integration with memristive crossbar arrays. The design is capable of driving 1T1R or 1S1R arrays for implementation of spiking neuromorphic systems, using a digital pulse-width modulation-based input scheme as well as an integrate-and-fire circuit connected to a digital counter for the output scheme. Furthermore, the controller design can be tiled to drive arbitrarily large arrays and flexibly assigned to different layers in a multilayer neural network design. The chips can be integrated on a PCBor in-package and utilized by an off-chip processor or FPGA.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Dec 01, 2021
Accession Number
AD1154062

Entities

People

  • Brady Taylor
  • Hai Helen Li

Organizations

  • Duke University

Tags

Communities of Interest

  • Advanced Electronics
  • Energy and Power Technologies

DTIC Thesaurus Topics

  • Abstracts
  • Air Force
  • Air Force Research Laboratories
  • Amplifiers
  • Artificial Intelligence
  • Artificial Intelligence Software
  • Circuits
  • Complementary Metal-Oxide Semiconductors
  • Computational Science
  • Computer Programming
  • Computer-Aided Design
  • Computers
  • Current Amplifiers
  • Energy Consumption
  • Engineering
  • Field Programmable Gate Arrays
  • Frequency
  • Integrated Circuits
  • Machine Learning
  • Memristors
  • Neural Networks
  • Quantum Tunneling
  • Semiconductors
  • Simulations
  • United States
  • Waveforms

Readers

  • Integrated Circuit Design and Technology.
  • Neural Network Machine Learning.

Technology Areas

  • AI & ML